CN113616188B - Respiration monitoring method based on frequency modulated continuous waves of inaudible sound - Google Patents
Respiration monitoring method based on frequency modulated continuous waves of inaudible sound Download PDFInfo
- Publication number
- CN113616188B CN113616188B CN202110918251.8A CN202110918251A CN113616188B CN 113616188 B CN113616188 B CN 113616188B CN 202110918251 A CN202110918251 A CN 202110918251A CN 113616188 B CN113616188 B CN 113616188B
- Authority
- CN
- China
- Prior art keywords
- signal
- frequency
- respiration
- signals
- respiratory
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000029058 respiratory gaseous exchange Effects 0.000 title claims abstract description 134
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000012544 monitoring process Methods 0.000 title claims abstract description 56
- 238000004458 analytical method Methods 0.000 claims abstract description 35
- 230000008859 change Effects 0.000 claims abstract description 19
- 230000005540 biological transmission Effects 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 238000002156 mixing Methods 0.000 claims abstract description 10
- 230000008030 elimination Effects 0.000 claims abstract description 9
- 238000003379 elimination reaction Methods 0.000 claims abstract description 9
- 230000000241 respiratory effect Effects 0.000 claims description 62
- 238000004422 calculation algorithm Methods 0.000 claims description 58
- 230000008569 process Effects 0.000 claims description 23
- 230000002159 abnormal effect Effects 0.000 claims description 18
- 230000036387 respiratory rate Effects 0.000 claims description 16
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 12
- 230000011218 segmentation Effects 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 8
- 210000000115 thoracic cavity Anatomy 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 6
- 108010076504 Protein Sorting Signals Proteins 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000012905 input function Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 2
- 239000012634 fragment Substances 0.000 claims 2
- 210000000038 chest Anatomy 0.000 abstract description 10
- 230000005236 sound signal Effects 0.000 abstract description 9
- 210000005069 ears Anatomy 0.000 abstract description 3
- 230000005237 high-frequency sound signal Effects 0.000 abstract description 3
- 238000004364 calculation method Methods 0.000 abstract description 2
- 238000002592 echocardiography Methods 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 description 18
- 238000002474 experimental method Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 13
- 238000000605 extraction Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 6
- 238000001035 drying Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000035565 breathing frequency Effects 0.000 description 4
- 201000010099 disease Diseases 0.000 description 4
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 4
- 230000036541 health Effects 0.000 description 4
- 230000033001 locomotion Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 208000000059 Dyspnea Diseases 0.000 description 3
- 206010013975 Dyspnoeas Diseases 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 2
- 241000282412 Homo Species 0.000 description 2
- 238000007791 dehumidification Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 230000037081 physical activity Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 206010020772 Hypertension Diseases 0.000 description 1
- 206010021079 Hypopnoea Diseases 0.000 description 1
- 208000020128 Mitral stenosis Diseases 0.000 description 1
- 206010027727 Mitral valve incompetence Diseases 0.000 description 1
- 208000025747 Rheumatic disease Diseases 0.000 description 1
- 206010044302 Tracheitis Diseases 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 206010002906 aortic stenosis Diseases 0.000 description 1
- 201000002064 aortic valve insufficiency Diseases 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000000090 biomarker Substances 0.000 description 1
- 229910002092 carbon dioxide Inorganic materials 0.000 description 1
- 239000001569 carbon dioxide Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 208000016569 congenital mitral valve insufficiency Diseases 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 208000019622 heart disease Diseases 0.000 description 1
- 208000018578 heart valve disease Diseases 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000004060 metabolic process Effects 0.000 description 1
- 208000005907 mitral valve insufficiency Diseases 0.000 description 1
- 208000006887 mitral valve stenosis Diseases 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000036651 mood Effects 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 239000002574 poison Substances 0.000 description 1
- 231100000614 poison Toxicity 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000552 rheumatic effect Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 208000008203 tachypnea Diseases 0.000 description 1
- 206010043089 tachypnoea Diseases 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/08—Detecting, measuring or recording devices for evaluating the respiratory organs
- A61B5/0816—Measuring devices for examining respiratory frequency
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Pulmonology (AREA)
- Physics & Mathematics (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Physiology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a respiration monitoring method based on inaudible sound frequency modulation continuous waves, which comprises the following steps: transmitting the modulated transmission signal within the human monitoring range such that the transmission signal is capable of being reflected by the thorax; receiving the reflected receiving signals, and preprocessing the receiving signals and the transmitting signals to obtain the receiving signals and the transmitting signals after noise elimination; mixing the preprocessed received signals and the preprocessed transmitted signals to obtain difference frequency signals; and (3) performing signal processing on the difference frequency signal, extracting a respiration signal from the change of the difference frequency signal, and monitoring the respiration of the human body. A respiration monitoring method based on inaudible sound frequency modulation continuous waves and unsupervised respiration rate calculation based on slow feature analysis. According to the method, extra hardware configuration is not needed, only a loudspeaker and a microphone which are arranged in the smart phone are used, high-frequency sound signals which cannot be sensed by human ears are used, and the human body respiration monitoring is completed by using the sound signal echoes, so that the method has robustness to different indoor environments.
Description
Technical Field
The invention relates to the technical field of medical equipment, in particular to a respiration monitoring method based on inaudible sound frequency modulation continuous waves.
Background
With the application and popularization of mobile terminal devices, the mobile communication industry will enter the true mobile information age. The mobile terminal not only has communication capability, but also becomes a comprehensive information processing platform and is more and more intelligent. A smart phone is one type of mobile terminal. It is a microcomputer integrating CPU, memory, solidified storage medium, input device and output device. The input and output devices comprise various sensors and functional hardware devices, and the sensors built in the smart phone can receive a lot of external information and complete processing of the information so as to extract more useful information, thereby performing more intelligent interaction. In the text, a loudspeaker in the smart phone is used for emitting modulated sound signals and a microphone is used for receiving echo signals, because the breathing of a human body can cause the fluctuation of the thorax, the motion can affect the emitted signals and change the propagation path of the signals, the received echo signals are analyzed and processed, the corresponding relation between the change of the propagation path and a breathing model is analyzed, the breathing signals are finally extracted, and the perception of the breathing of the human body is completed.
The intelligent terminal device is used for sensing the respiration, and the mechanism of the respiration is firstly known. During metabolism, the body needs to continuously suck oxygen from the outside to discharge carbon dioxide, and the gas exchange between the body and the environment is called breathing. Respiration can inform the health condition of a human body, and is one of the most important vital signs of the human body. The frequency, rhythm and depth of breathing can be changed by the influence of diseases, poisons or drugs. Adults breathe more than 24 times per minute, known as breathlessness or breathlessness; breaths per minute are less than 10, called hypopneas. For example, abnormal heart breathing frequency, rheumatic valvular heart disease (mitral stenosis and insufficiency, aortic stenosis and insufficiency, etc.), hypertension, heart disease, tracheitis, etc. all cause abnormal breathing frequency, which affects people's life and even threatens life. The breathing rate is of great importance as a marker reflecting health and safety issues in humans. By means of the respiration rate, whether the body of the subject is in a healthy state or not and whether the body of the subject is threatened by diseases or not can be judged. The elderly are more in need of care of their family as a high-incidence population of these diseases. Particularly, the elderly living alone will suffer from diseases during sleep, and if the elderly cannot be found and treated immediately, irreparable results will be produced. The main content of the study herein is focused on using a smartphone to enable breath monitoring indoors.
Disclosure of Invention
In view of the above-mentioned drawbacks and deficiencies, it is an object of the present invention to provide a method for monitoring respiration based on inaudible sound fm continuous waves.
In order to achieve the above purpose, the technical scheme of the invention is as follows:
a respiration monitoring method based on frequency modulated continuous waves of inaudible sounds comprises the following steps:
1) Transmitting the modulated transmission signal within a human monitoring range such that the transmission signal is capable of being reflected by the thoracic cavity;
2) Receiving the reflected receiving signals, and preprocessing the receiving signals and the transmitting signals to obtain the receiving signals and the transmitting signals after noise elimination;
3) Mixing the preprocessed received signal and the preprocessed transmitted signal to obtain a difference frequency signal;
4) And carrying out signal processing on the difference frequency signal, extracting a respiration signal from the change of the difference frequency signal, and monitoring the respiration of the human body.
The transmitting signal is a high-frequency chirp signal of 18-20 kHZ.
The step 2) of preprocessing the received signal and the transmission signal to obtain the received signal and the transmission signal after noise elimination, specifically comprising:
2.1, carrying out short-time Fourier transform on the received signal to obtain a time-frequency sequence of the received signal;
2.2, traversing the whole time-frequency sequence until finding a point with the first frequency not equal to 0 in the sequence, and recording the point as an anchor point;
2.3, finishing traversing to obtain a preprocessed receiving signal;
2.4, preprocessing the transmitting signal, and obtaining a time-frequency sequence of the transmitting signal through short-time Fourier transform; traversing the whole time-frequency sequence until finding a point with the first frequency not equal to 0 in the sequence, and recording as an anchor point;
and 2.5, finishing traversing to obtain a preprocessed transmitting signal.
The step 3) of mixing the preprocessed signals with the transmitted signals to obtain difference frequency signals specifically includes:
3.1, mixing the data between the transmission signal and the receiving signal after the preprocessing and from the anchor point to the signal end point to obtain a mixed signal;
3.2, selecting a Butterworth low-pass filter with the cutoff frequency of 1kHZ to eliminate high-frequency components in the mixed signal, and realizing low-pass filtering processing;
and 3.3, calculating the frequency difference between the transmitting signal and the receiving signal after the low-pass filtering processing to obtain a difference frequency signal.
The step 4) of performing signal processing on the difference frequency signal and extracting the respiration signal from the change of the difference frequency signal specifically includes:
4.1, calculating a PSD signal based on the difference frequency signal:
P＝k|S(f)| 2 wherein, in the process,
in the formula, S (f) is a difference frequency signal, S (t) is a sequence after short-time Fourier transform, k is a scalar with a real value, wherein fs is a sampling frequency, w (n) is a window function, and L is the number of windows set when S (t) realizes short-time Fourier transform;
4.2, performing drying treatment on the PSD signal;
4.3, segmenting the PSD signal after the drying treatment, inputting the segmented PSD signal into a slow feature analysis algorithm, and extracting a respiratory signal from an environment background;
4.4, detecting the number of peak values of the respiratory waveforms in the respiratory signals by using a peak value monitoring algorithm, calculating the respiratory rate and realizing the monitoring of the human respiration.
The drying processing of the PSD signal specifically includes:
and through an adaptive signal normalization algorithm, the noise of the PSD signal caused by the self interference of the difference frequency signal is eliminated. The specific process is as follows:
(1) A sliding window with the length of l is set to calculate the standard deviation of the difference frequency signal, and a standard deviation sequence s _ std = { s _ std1, s _ std2, \ 8230; \8230;, s _ stdn } is obtained. The length of l is less than a sweep period of the chirp signal, and the operation aims to reflect the volatility of the PSD signal caused by invalid signals;
(2) Then, the std sequence is traversed, and the starting point and the ending point of the invalid signal are identified by adopting a dynamic threshold method. Through a plurality of experiments: finally, a length l is set 1 Calculating the mean value based on the sliding window signal as the dynamic threshold value
And realizing the identification and elimination of invalid signals.
(3) The threshold is determined immediately after the start and end points of the invalid signal are identified. First, a starting point of the invalid signal is determined.
The PSD signal after the drying treatment is segmented and then input into a slow feature analysis algorithm, and a respiratory signal is extracted from an environment background, and the method specifically comprises the following steps:
4.3.1, dividing the input PSD signal into a plurality of equal-length segments with the lengths of n sampling points to finish first division;
4.3.2, each equal-length segment is re-segmented by using a time delay-based segmentation method, unit time delay is increased in an incremental mode for each equal-length segment, and each re-segmentation has an overlapping part;
4.3.3, finally, dividing one column of delta-dimensional PSD data N into c columns of M-dimensional PSD matrixes M; n, M is represented as:
where δ represents the delay of δ sample points.
And 4.3.4, inputting the matrix serving as an input signal into a slow characteristic analysis algorithm, and extracting a respiratory signal.
The matrix is input into a slow characteristic analysis algorithm as an input signal to extract a respiratory signal, and the principle is as follows:
the slow feature analysis algorithm is divided into a learning process (or called training process) and a testing process. In this context, however, a learning process is implemented, in which a slowly varying breathing signal is extracted from the received echo signal. In the learning process of slow feature analysis, it is assumed that the input signal and the output signal are x (t) and y (t), respectively. Suppose the input signal is x (t) = [ x = [) 1 (t),......,x I (t)] T Output signal y (t) = [ y = [) 1 (t),......,y J (t)] T . The objective is to find an input-output function g (x) = [ g = [) 1 (x),......,g J (x)] T Generating a J-dimensional inputAnd (6) outputting a signal. The output signal y (t) is a matrix, and the optimal result needs to be screened from the sequence, and the optimal result is assumed to be the jth component y in y (t) j (t) of (d). Then y is j (t) satisfies: delta (y) j (t))＝<y′j(t)> 2 Minimum, y' j (t) is the first derivative of y (t). After the target expression is determined, a constraint condition is added to the expression, such as formula (2-7) -formula (2-9).
<y j >＝0 (2-7)
In formulae (2-7) to (2-9)
The constraint conditions (2-7) are used for meeting the requirements of the constraint conditions (2-8) and the constraint conditions (2-9); the constraint conditions (2-7) have the function of ensuring that the output signal carries useful information, and avoiding that the output signal becomes constant by some simple solutions; the constraints (2-9) represent decorrelation such that the information carried by the components does not overlap with each other. Calculated, each dimension y of the output signal j All representing different types of information in the signal, and the constraints (2-8) are such that y is in the output signal 1 Is the slowest varying, y 2 Degree of slowness of change of (a) secondly, y 3 And so on for the next time.
The method specifically comprises the following steps:
inputting the data of the matrix into the SFA to obtain output signals y (t), wherein the result in y (t) is in accordance with delta (y) j (t)) arranging from small to large, and selecting a first row of y (t) as an optimal solution; if the optimal solution in y (t) can not reach the expected requirement, taking y (t) as the input function x (t), and repeating the steps until delta (y) is twice j (t)) is less than 0.01. The above-mentionedThe method comprises the following steps of detecting the number of peak values of respiratory waveforms in respiratory signals by using a peak value monitoring algorithm, and calculating the respiratory rate, wherein the method specifically comprises the following steps:
4.4.1, setting the length wn of the sliding window, the length wnb of the back window and the respiration signal sequence B (t), 4.4.2, and calculating the maximum value of the sliding window to be P and the maximum value of the back window to be S.
4.4.3, judging whether an abnormal value appears in the two windows or not, and if the abnormal value appears, eliminating the abnormal value. If there is no abnormal value, continue.
4.4.4, if the peak of P is less than the peak of S and the separation between the two points is less than one breath interval, then S is considered a breath peak. If the distance between the two points is larger than one respiration interval, P is regarded as a respiration peak, S is regarded as P, and the step 4.4.3 is carried out. If the peak value of P is greater than S, the window is moved backward and the process continues at step 4.4.3.
4.4.5, do secondary scan, will not be in the threshold interval [ thr 1 ,thr 2 ]The peak value of (2) is deleted. thr 1 ＝μ result -σ res ；thr 2 ＝μ result +σ result //μ result Is the result mean, δ result Is the result standard deviation.
4.4.6, the exact number of peaks in the respiratory signal bn is obtained.
The calculated respiratory rate is:
the peak value number n is the number of breaths, and the breathing rate v is calculated according to the relation between the number of breaths n and the length fn of the data of the first equal-length division b ：
Wherein fs is the sampling frequency. v. of b Representing the breath rate in minutes in bpm. bpm represents the number of breaths per minute; according to the respiratory rate v b And (4) calculating the number of breaths per minute, namely the breathing rate, and judging whether the abnormal breathing condition exists or not by using a formula.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a respiration monitoring method based on inaudible sound frequency modulation continuous waves, which is characterized in that the respiration of a human body is slowly changed relative to a transmitting signal by using a high-frequency sound signal which cannot be sensed by human ears only by using a built-in loudspeaker and a built-in microphone of a smart phone without additional hardware configuration and by using a sound signal echo. According to the characteristic, an unsupervised slow characteristic analysis algorithm is introduced to extract a respiratory signal with a relatively microscale from a rapidly-changing signal, so that the aim of monitoring the respiration indoors by using a smart phone is fulfilled.
Furthermore, the invention provides a self-adaptive signal normalization algorithm to eliminate the self-interference of the difference frequency signal and reduce the error in the process of extracting the respiratory signal.
Further, the present invention uses inaudible sound signals having a frequency much greater than the breathing frequency of the human body, which is a slowly varying feature relative to the emitted signal. According to the characteristic, an unsupervised slow feature analysis algorithm is introduced to extract a respiratory signal with relative micro-scale from a rapidly changing signal.
Furthermore, according to the characteristic that the respiration signal extracted by the slow characteristic analysis algorithm reflects the respiration change through the peak value change, a peak value detection algorithm is designed, the occurrence of wrong peak values is reduced, and the accuracy of the respiration monitoring result is improved.
Drawings
FIG. 1 is a flow chart of a method for monitoring respiration based on an inaudible acoustic FM continuous wave in accordance with the present invention;
FIG. 2 is a diagram of a frequency modulated continuous wave difference frequency signal of the present invention; wherein (a) is a plot of the transmitted signal and the received signal, and (b) is a plot of the difference frequency signal;
FIG. 3 is a diagram of the pre-dehumidification difference frequency signal of the present invention;
FIG. 4 is a diagram of a PSD signal based on a difference frequency signal according to the present invention;
FIG. 5 is a schematic diagram of noise start and end point identification according to the present invention;
FIG. 6 is a time delay segmentation chart of the present invention;
FIG. 7 is a diagram of the feature extraction process of the present invention;
FIG. 8 is a graph of standard peak detection results;
FIG. 9 is a graph of peak detection results for the present invention;
FIG. 10 is a graph of the effect detection results of different distances according to the present invention;
FIG. 11 is a graph of the impact detection results of different angles according to the present invention;
FIG. 12 is a graph of the results of impact detection for different environments in accordance with the present invention;
fig. 13 is a diagram of the impact detection results of different users of the present invention.
Detailed Description
The present invention will now be described in detail with reference to the drawings, wherein the described embodiments are only some, but not all embodiments of the invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, belong to the scope of the present invention.
And (4) monitoring the respiration aiming at different scenes and different crowds. In addition to the work of complex or special scenarios, the challenge of respiratory monitoring in general indoor environments is generally the following:
breathing is a minute movement. The invention relates to research on non-contact respiration monitoring based on the observation that respiration can cause thoracic cavity fluctuation. However, the amplitude of the thorax undulations is small compared to other physical activities, about 1cm-2cm, which presents a huge challenge to the extraction of the breathing signal.
And (4) accuracy. Since respiration is one of the ultimate biological markers reflecting human health and safety issues, health and safety concerns are one of the focuses of attention. Therefore, the accuracy of the respiratory signal extraction and respiratory rate calculation is also a goal pursued by respiratory monitoring related efforts. Respiratory monitoring is robust in different environments. The aim of the invention is to achieve an accurate monitoring of the breathing rate using commercially available acoustic devices. A high-frequency chirp signal of 18-20 kHZ is transmitted by adopting an FMCW technology based on sound signals. The signal of the frequency band can be heard by only a few people, and discomfort cannot be brought to the monitored signal.
Specifically, as shown in fig. 1, the present invention provides a respiration monitoring method based on inaudible sound frequency modulated continuous wave, including:
1) Transmitting the modulated transmission signal within the human body monitoring range such that the transmission signal is capable of being reflected through the thoracic cavity;
the invention mainly collects three kinds of data: a transmitted signal, a received signal and real data. When monitoring respiration, a user holds a mobile phone by hand and monitors the position at a certain distance from the thoracic cavity, controls a built-in loudspeaker and a built-in microphone of the mobile phone by using a mobile phone APP, and stores a transmitting signal and a receiving signal recorded by the microphone as a PCM file. Meanwhile, a 3-lead ECG monitor HealForce PC-80B is used as real data.
In addition, by adopting FMCW technology based on sound signals, high-frequency chirp signals of 18-20 kHZ are transmitted, and the frequency sweep period is 0.08s. The signal of the frequency band can be heard only by a few people, and discomfort can not be brought to the monitored object.
2) Receiving the reflected receiving signals, and preprocessing the receiving signals and the transmitting signals to obtain the receiving signals and the transmitting signals after noise elimination;
2.1, carrying out short-time Fourier transform on the received signal to obtain a time-frequency sequence of the received signal;
2.2, traversing the whole time-frequency sequence until finding a point with the first frequency not equal to 0 in the sequence, and recording the point as an anchor point;
2.3, finishing traversing to obtain a preprocessed receiving signal;
2.4, preprocessing the transmission signal, and obtaining a time-frequency sequence of the transmission signal through short-time Fourier transform; traversing the whole time-frequency sequence until finding a point with the first frequency not equal to 0 in the sequence, and recording as an anchor point;
and 2.5, ending traversal to obtain the preprocessed transmitting signal.
With the program controlled opening and closing of the speaker and microphone, the microphone does not start collecting data immediately but rather with a short delay as the program controlled microphone is activated. In this section, a naive frequency detection method is used to eliminate null data due to delay. The specific process is as follows: firstly, a Time-frequency sequence of a received signal is obtained through Short-Time Fourier Transform (STFT), then the whole Time-frequency sequence is traversed until a point with the first frequency not equal to 0 in the sequence is found and is marked as an anchor point, and the traversal is finished. And then, applying STFT to the transmitting signal, deleting data between the transmitting signal and the receiving signal from the first sampling point to the anchor point, and inputting data between the transmitting signal and the receiving signal from the anchor point to the signal end point to the next step for mixing.
3) Mixing the preprocessed received signal and the preprocessed transmitted signal to obtain a difference frequency signal;
3.1, mixing the data between the transmission signal and the receiving signal after the preprocessing and the data between the signal from the anchor point to the signal end point to obtain a mixed signal;
3.2, selecting a Butterworth low-pass filter with the cutoff frequency of 1kHZ to eliminate high-frequency components in the mixed signal and realize low-pass filtering processing;
and 3.3, calculating the frequency difference between the transmitting signal and the receiving signal after the low-pass filtering processing to obtain a difference frequency signal.
The difference frequency signal comprises a valid signal S E And an invalid signal S I Two parts, calculating the difference frequency signal f of the transmitting signal from the first sweep frequency period T by formula b ：
Where n is the nth sweep period of the transmitted signal, n =1,2,3, \8230;, T is the sweep period, v is the speed of sound, and d is the distance between the device and the reflector.
The most intuitive expression of breathing in humans is the fluctuation of the thorax. When a user holds the mobile phone to monitor respiration at a position away from the chest, the mobile phone transmits a modulated chirp signal through a loudspeaker, as shown in fig. 2 (a), the first signal is the frequency of the transmitted signal, and the signal is reflected by the chest and then collected by a microphone. Therefore, it is possible to track changes in the signal path due to fluctuations in the thorax caused by breathing and extract the respiratory signal from the changes. The signal path changes, namely the frequency difference between the transmitting signal and the receiving signal changes, and the breathing signal is extracted from the change of the frequency difference.
According to the distance and difference frequency signal relation formula derived from C-FMCW 1,
it can be known that the frequency difference between the transmitting signal and the receiving signal is in a direct proportion relation with the distance between the chest cavity and the smart phone. Therefore, the purpose of monitoring respiration can be achieved by tracking the difference frequency signal between the transmitted signal and the received signal, and the difference frequency is called the difference frequency signal.
The distance is in a direct proportion relation with the difference frequency signal, the true distance between the chest cavity and the smart phone does not need to be estimated, and the respiratory signal is extracted from the change of the difference frequency signal by tracking the change of the difference frequency signal. The present invention captures the variation of the frequency of the difference frequency signal using an energy spectral density signal based on the difference frequency signal. Although the change of the frequency of the difference frequency signal can be captured by the energy spectrum density signal, the difference frequency signal has interference, and the interference can cause interference on the respiratory signal which is extracted by using a slow feature analysis algorithm. Therefore, the variation of the signal path caused by the thorax fluctuation caused by respiration of the difference frequency signal is analyzed by analog data, and the respiration signal is extracted according to the variation.
Observing the difference frequency signal shown in fig. 2 (b) (the lower line represents the frequency of the difference frequency signal calculated from a closer reflector, and the upper line represents the frequency of the difference frequency signal calculated from a farther reflector), the signal in the black box in fig. 2 (b) is considered herein as "noise", i.e., self-interference of the difference frequency signal. The relationship of "noise" to distance is the inverse of the representation of the model. "noise" is not used for the extraction of the breathing signal. After a plurality of times of simulation experiments, the difference frequency signals obtained from the reflected signals of objects with different distances have the noise, and cannot be filtered by a filter. Therefore, the invention provides a signal normalization algorithm to further eliminate noise of the signal. After noise is removed from the difference frequency signal, a PSD signal is calculated and used as input data of a slow feature analysis algorithm to extract a respiratory signal.
4) And carrying out signal processing on the difference frequency signal, extracting a respiration signal from the change of the difference frequency signal, and monitoring the respiration of the human body.
The method specifically comprises the following steps:
4.1, calculating a PSD signal based on the difference frequency signal:
P＝k|S(f)| 2 wherein
in the formula, S (f) is a difference frequency signal, S (t) is a sequence after short-time Fourier transform, k is a scalar with a real value, wherein fs is a sampling frequency, w (n) is a window function, and L is the number of windows set when S (t) realizes short-time Fourier transform;
4.2, performing dehumidification treatment on the PSD signal;
the PSD signal is calculated from data collected in a real indoor environment as shown in fig. 3. After multiple experiments, the periodic signal notch condition appears in the spectrogram of the difference frequency signal and the waveform of the PSD signal drawn in each experiment, and the notched part of the signal, the noise, namely the self interference of the difference frequency signal can be determined according to the comparison of the chirp signal parameter and the figure value. In order to eliminate the interference, the invention provides an adaptive signal normalization algorithm, which eliminates the noise of the PSD signal caused by the self-interference of the difference frequency signal.
As shown in fig. 3, the "noise" is regularly distributed in the difference frequency signal, and the PSD amplitude is lower than the effective part, that is, when the "noise" is processed, the signal has obvious fluctuation, so that the identification and subtraction of the "noise" are realized by using a method based on the standard deviation of the signal. The specific implementation process of the algorithm comprises the following steps:
(1) A sliding window with the length of l is set to calculate the standard deviation of the difference frequency signal, and a standard deviation sequence s _ std = { s _ std1, s _ std2, \ 8230; \8230;, s _ stdn } is obtained. The length of l is less than a sweep period of the chirp signal, and the operation aims to reflect the volatility of the PSD signal caused by invalid signals;
(2) Then, the std sequence is traversed, and the starting point and the ending point of the invalid signal are identified by adopting a dynamic threshold method. Through a plurality of experiments: finally, a length l is set 1 Calculating a mean value based on the sliding window signal as a dynamic threshold value
And realizing the identification and elimination of invalid signals.
(3) The threshold is determined immediately after the start and end points of the invalid signal are identified. First, a starting point of the invalid signal is determined.
(4) And finally, determining the end point of the invalid signal, observing that a point D in the graph 5 is a judgment point which accords with the end position of the invalid signal in the standard deviation sequence, and having a difference with the position of an E point of the end point of the actual invalid signal. In order to recognize "noise" to the maximum, the position of the point D plus the empirical error η is determined as the end point E of "noise". And determining a starting point and an end point, and directly subtracting the invalid signal in the traversal process.
The signal normalization algorithm eliminates the noise in the difference frequency signal, the length of the difference frequency signal is reduced, and the subsequent peak value detection algorithm detects the respiratory wave peak by using the length of the respiratory interval as a threshold value, so that the PSD signal is subjected to difference by adopting an interpolation algorithm. The data generated by the difference is input to the next step.
4.3, segmenting the PSD signal after the drying treatment, inputting the segmented PSD signal into a slow feature analysis algorithm, and extracting a respiratory signal from an environment background;
the breathing monitoring for the user is realized for a long time, the sampling rate of the acoustic equipment in the mobile phone is 48kHZ, and the processed data volume is very large. Therefore, the input data needs to be segmented, which can reduce the computation overhead, increase the computation speed, and has relatively low requirement on hardware. In image processing, the data input into the SFA algorithm is segmented by using a frame difference method. A PSD-based signal partitioning method is used in the present invention. The segmentation algorithm mainly comprises two steps of equal-length segmentation and time-delay segmentation.
The PSD signal after the drying treatment is segmented and then input into a slow feature analysis algorithm, and a respiratory signal is extracted from an environment background, and the method specifically comprises the following steps:
4.3.1, dividing the input PSD signal into a plurality of equal-length segments with the lengths of n sampling points to finish first division;
4.3.2, each equal-length segment is re-segmented by using a time delay-based segmentation method, unit time delay is increased in an incremental mode for each equal-length segment, and each re-segmentation has an overlapping part; the specific segmentation process is shown in fig. 6:
4.3.3, finally, dividing a column of delta-dimensional PSD data N into c columns of M-dimensional PSD matrixes M; n, M is represented as:
where δ represents the delay of δ sample points.
And 4.3.4, inputting the matrix into a slow characteristic analysis algorithm as an input signal, and extracting a respiratory signal.
The matrix is input into a slow characteristic analysis algorithm as an input signal, and the extraction of the respiratory signal specifically comprises the following steps:
inputting the data of the matrix into the SFA to obtain output signals y (t), wherein the result in y (t) is in accordance with delta (y) j (t)) arranging from small to large, and selecting a first row of y (t) as an optimal solution; if the optimal solution in y (t) can not reach the expected requirement, taking y (t) as the input function x (t), and repeating the steps until delta (y) is twice j (t)) is less than 0.01.
The slow feature analysis algorithm is classified as learnedA program (or called a training process) and a test process. In this context, however, a learning process is implemented to extract a slowly varying breathing signal from the received echo signal. In the learning process of slow feature analysis, it is assumed that the input signal and the output signal are x (t) and y (t), respectively. Suppose the input signal is x (t) = [ x = [) 1 (t),......,x I (t)] T Output signal y (t) = [ y = [) 1 (t),......,y J (t)] T . The objective is to find an input-output function g (x) = [ g = [) 1 (x),......,g J (x)] T And generating a J-dimensional output signal. The output signal y (t) is a matrix, and the optimal result needs to be screened from the sequence, assuming that the optimal result is the jth component y in y (t) j (t) of (d). Then y is j (t) satisfies: delta (y) j (t))＝<y′ j (t)> 2 Minimum, y' j (t) is the first derivative of y (t). After determining the target expression, adding a constraint condition to the expression, as follows:
<y j >＝0 (2-7)
wherein,
the constraint conditions (2-7) are used for meeting the requirements of the constraint conditions (2-8) and the constraint conditions (2-9); the constraint conditions (2-7) have the function of ensuring that the output signal carries useful information, and avoiding that the output signal becomes constant by some simple solutions; the constraints (2-9) represent decorrelation such that the information carried by the components does not overlap with each other. Each dimension yj of the output signal is calculated to represent a different type of information in the signal, and the constraints (2-8) are such that y1 changes most slowly and y2 changes slowly in the output signalSecond, y3 is the next and so on.
The PSD sequence is segmented as described above, and the matrix is input as an input signal to a slow signature analysis algorithm. As shown in fig. 7, the process of extracting the respiration signal is described. The use of a slow feature analysis algorithm to extract the respiratory signal is due to two reasons: first, a slow feature analysis algorithm distinguishes the respiratory signal from the background. In the process of extracting the respiratory signal, the respiratory signal has the advantages of periodicity, small motion amplitude, weak reflection energy and the like, and great challenges are brought to the extraction of the respiratory signal. Periodic breathing signals can be distinguished from background using a slow signature analysis algorithm. The use of unsupervised slow feature analysis algorithms to extract human body activity from the background has the disadvantage that the differentiation of various actions is not ideal, but does not affect the system, since the invention aims to extract a signal of breathing. Secondly, the unsupervised slow feature analysis algorithm does not need to extract respiratory features in advance, and a respiratory feature database is built. The scheme does not need offline feature acquisition, and reduces labor cost.
4.4, detecting the number of peak values of the respiratory waveforms in the respiratory signals by using a peak value monitoring algorithm, calculating the respiratory rate and realizing the monitoring of the human respiration.
A standard peak detection algorithm may determine the transition point of the signal from an increasing trend to a decreasing trend. However, this can result in many false peaks in the respiratory chest movement signal, as shown in FIG. 6, and many unexpected peaks are detected. Therefore, a peak detection algorithm is proposed, whose targets are two: firstly, reducing the occurrence of wrong peak values in the detection process; second, a peak can be accurately detected.
The method for detecting the number of peaks of the respiratory waveform in the respiratory signal by using the peak monitoring algorithm and calculating the respiratory rate specifically comprises the following steps:
4.4.1, setting a sliding window length wn, a back window length wnb, and a breathing signal sequence B (t),
4.4.2, calculating the maximum value of the sliding window as P, and the maximum value of the rear window as S.
4.4.3, judging whether abnormal values appear in the two windows or not, if the abnormal values appear, eliminating the abnormal values, and if the abnormal values do not appear, continuing;
4.4.4, if the peak value of P is smaller than the peak value of S and the distance between the two points is smaller than one breathing interval, the S is regarded as a breathing peak, if the distance between the two points is larger than one breathing interval, the P is regarded as a breathing peak, the S is regarded as the P, and the step 4.4.3 is carried out. If the peak value of P is greater than S, the window is moved backward and the process continues at step 4.4.3.
4.4.5, performing secondary scanning, and not in the threshold interval [ thr1, thr2 ]]Deleting the wave peak value; wherein, the flow rate of the water is controlled by the control unit. thr 1 ＝μ result -σ result ；thr 2 ＝μ result +σ result //μ result Is the result mean value, δ result Is the result standard deviation.
4.4.6, get the exact number of peaks bn in the respiration signal.
The calculated respiratory rate is:
and calculating the number of respiratory waveform peaks of the respiratory signal sequence by the peak detection algorithm. The results of the detection are shown in FIG. 7. After the respiratory signal is extracted, peak detection is carried out on the respiratory signal extracted by the slow characteristic analysis algorithm through a peak detection algorithm, the peak number n is the respiratory number, and the respiratory rate v is calculated according to the relationship between the respiratory number n and the data length fn of the first equal-length segmentation b ：
Wherein fs is the sampling frequency. v. of b Representing the breath rate in minutes in bpm. bpm represents the number of breaths per minute; according to the respiration rate v b And calculating the number of breaths per minute, namely the breathing rate, and judging whether the abnormal breathing condition exists or not by using a formula. The respiration rate is finally estimated to be 17.39bpm through the experimental result, and the respiration is normal. When the respiratory rate is less than 10bpm, the breath is considered to be slow; when the breathing rate exceeds 24bpm, breathlessness is considered. And both the slow breathing and the tachypnea are judged as abnormal breathing.
The invention firstly provides an unsupervised respiration monitoring system based on inaudible sound. The system does not need additional hardware configuration, only uses a built-in loudspeaker and a built-in microphone of the smart phone, uses high-frequency sound signals which cannot be sensed by human ears, and uses sound signal echoes to complete human body respiration monitoring.
Secondly, a method for eliminating self-interference of the difference frequency signal based on signal normalization is provided. And eliminating high-frequency noise in the difference frequency signal by using the jump characteristic from low frequency to high frequency in the difference frequency signal.
And thirdly, providing an unsupervised respiratory signal feature extraction method based on slow feature analysis. And performing equal-length segmentation and delayed segmentation on the one-dimensional energy spectrum density signal based on the difference frequency signal, and taking the one-dimensional energy spectrum density signal as input data of a slow characteristic analysis algorithm to finally extract a respiratory signal.
Finally, 12 volunteers were breath monitored in three indoor environments to evaluate system performance. Experimental results show that the proposed method achieves a median error of less than 2 times/min. The method can realize effective respiration monitoring for different users, environmental noise and positions (within 0.4 m).
In the preliminary verification experiment, the difference frequency signal has a component with higher frequency, and the signal of the high-frequency component can not be really used for extracting the respiratory signal, namely the difference frequency signal has interference in the respiratory signal extraction process. Therefore, the self-interference of the difference frequency signal is eliminated by the self-adaptive signal warping algorithm, and the error in the respiratory signal extraction process is reduced.
The invention uses inaudible sound signals at frequencies much greater than the human breathing, a slowly varying characteristic of human breathing relative to the transmitted signal. Based on this feature, an unsupervised slow feature analysis algorithm is introduced herein to extract a relatively micro-scale respiration signal from a rapidly varying signal.
According to the characteristic that the respiration signal extracted by the slow characteristic analysis algorithm reflects the respiration change from the peak value change, a peak value detection algorithm is designed, the occurrence of wrong peak values is reduced, and the accuracy of the respiration monitoring result is improved.
The invention has the following effects:
(1) In a preliminary verification experiment, it is found that a higher frequency component exists in the difference frequency signal, and the signal of the high frequency component cannot be really used for extracting the respiratory signal, namely the difference frequency signal itself has interference in the process of extracting the respiratory signal. Therefore, the invention provides a self-adaptive signal normalization algorithm to eliminate the self-interference of the difference frequency signal and reduce the error in the respiratory signal extraction process.
(2) As used herein, the frequency of an inaudible sound signal is much greater than the breathing frequency of the human body, which is a slowly varying feature relative to the emitted signal. Based on this feature, an unsupervised slow feature analysis algorithm is introduced herein to extract a relatively micro-scale respiration signal from the rapidly varying signal.
(3) According to the characteristic that the respiration signal extracted by the slow characteristic analysis algorithm reflects the respiration change from the peak value change, a peak value detection algorithm is designed, the occurrence of wrong peak values is reduced, and the accuracy of the respiration monitoring result is improved
The experimental results are as follows:
in order to verify the feasibility and robustness of the respiration monitoring method based on the frequency modulation continuous wave of the inaudible sound, a large number of experiments are formulated and implemented in the invention. 12 subjects participated in the experiment, including males and females between 20 and 30 years of age, and were between 164cm and 191cm in height. The 3-lead ECG monitor Heal Force PC-80B is used as real data herein. The sampling rate of this ECG monitor is 150Hz. The heart rate is detected here using ECG and the respiration rate is one quarter of the heart rate is used as the real data for experimental evaluation.
1. Influence of relative position:
in whole experimentation, discover human and smart mobile phone's relative position, also can cause the influence of different degree to the experimental result equally, follow the influence of different positions to breathing monitoring through the experiment volume observation. The location in the present invention is represented by the distance and angle between the target monitor and the intelligence.
Therefore, the distance and the angle are respectively tested, and the relationship between the different distances and the different angles and the respiratory rate estimation error is known through the test result. First, experimental evaluations were performed for different distances. Experiments of the user at different distances in a direction perpendicular to the human body. The results of the experiment are shown in FIG. 10.
The following results were obtained by observation: the respiratory rate estimation error increases with the distance between the target monitoring object and the smart phone, and when the distance is within 10cm and 10cm, the mean respiratory error value is less than 0.25bpm. When the distance is increased to 30cm, the median of the respiration rate estimation error is larger than 3bpm, and the respiration rate estimation error increase rate is accelerated. Acourdar proposes a Virtual Acoustic Beamforming Model (VABM). From this model, it can be concluded that the distance is inversely proportional to the sound intensity, and when the distance is small, the rate of decrease of the sound intensity is fast, and as the time increases, the rate of decrease of the sound intensity gradually slows down. As is the relationship of distance to breathing rate estimation error herein.
Next, the relationship of the angle of observation to the estimation error of the breathing rate estimation is experimentally observed. The invention tests three angles respectively. When the surface of the smart phone faces upwards, the position of the bottom loudspeaker facing the chest of the target monitor is expressed as 90 degrees, in the experiment, the smart phone rotates in the anticlockwise direction, and then respiratory rate estimation errors of 70 degrees and 35 degrees are monitored. The results are shown in FIG. 11.
The experimental result shows that when the angle between the smart phone and the target monitor is gradually increased and the respiration rate estimation is larger than 70 degrees, the average respiration rate estimation error is not more than 1.5bpm, but when the angle is increased again and the smart phone collects echo signals, most of the signals are reflected and cannot be received by the microphone, so the respiration rate estimation error is increased, the fluctuation is obvious and the error is large.
2. Influence of relative position:
indoor furnishings and human activities may affect user behavior and signal changes and, therefore, may affect the performance of the respiration monitoring methods presented herein. The present invention separately analyzes the user's respiratory data collected from different indoor environments. Three indoor environments of a kitchen, a bedroom and a living room: there is human activity in the kitchen, but there is less furnishing in the space; no other human activity is present in the bedroom than the user being monitored; the living room is the most complex environment, and various furniture and household appliances are more and more displayed indoors; in addition to the monitored user, the activities of other family members are an indoor environment with more complex noise.
Fig. 12 shows the respiratory rate estimation error of different indoor environments, and the experimental results show that: when the user is in an indoor environment (bedroom) where human activity is low, the median is 0.45bpm, although the resulting fluctuation of the respiration rate estimation error is large. Compared with the other two environments with human activities, the accuracy is higher. From the overall comparison results, in more complex indoor environments, human activity affects the respiration monitoring methods presented herein to a greater extent than the ambient noise in the room. Therefore, the method proposed herein may be effective in estimating the respiration rate when there is little other human activity or when the other user is a little further away from the monitored user.
3. Influence of different users:
in addition to the above-mentioned effects, the effects of different users should also be taken into account. In each set of comparison experiments, the respiration rate estimation error results are obtained when the same user performs respiration monitoring. In daily life, people wear different clothes, are in different indoor environments, have great differences in breathing characteristics, and compare the breathing rate estimation errors of different users in order to further verify the robustness of the breathing monitoring method provided by the text.
A total of 12 volunteers participated in the experiment, and each volunteer randomly selected different experimental environments (living room, kitchen and bedroom) to complete the experiment while wearing different clothes. Experimental data of 7 volunteers were randomly selected in this section, and the experimental results are plotted as shown in fig. 11.
By observing fig. 11, it is found that the breathing rate estimation error of the user 5 is the largest and the fluctuation range of the result is the widest. The method provided by the invention cannot correctly monitor the breathing condition of the patient. This is because, in the course of the experiment, the emotion of the user greatly fluctuates, which results in the variation range of the heartbeat rate of 30bpm, and the relationship between the respiration rate and the heartbeat rate is no longer linear. Besides the influence of the human body, the clothes of the user also has certain influence on the result. For example, the user 3 and the user 4 wear thicker sweaters and sweaters during experiments, the clothes not only reduce the displacement of the fluctuation of the chest cavity during monitoring respiration, but also have certain sound absorption effect and weaken the energy of signals. The breathing rate estimation error is around 4 bpm. The experimental results show that: the heartbeat rate has the greatest influence on the method provided by the text, and when the mood of the user fluctuates little, the breath of the user can be captured and the breath rate can be calculated.
It will be appreciated by those skilled in the art that the above embodiments are merely preferred embodiments of the invention, and thus, modifications and variations may be made in the invention by those skilled in the art, which will embody the principles of the invention and achieve the objects and objectives of the invention while remaining within the scope of the invention.
Claims (8)
1. A respiration monitoring method based on an inaudible sound frequency modulation continuous wave is characterized by comprising the following steps:
1) Transmitting the modulated transmission signal within a human monitoring range such that the transmission signal is capable of being reflected by the thoracic cavity;
2) Receiving the reflected receiving signals, and preprocessing the receiving signals and the transmitting signals to obtain the receiving signals and the transmitting signals after noise elimination;
3) Mixing the preprocessed received signals and the preprocessed transmitted signals to obtain difference frequency signals;
4) Carrying out signal processing on the difference frequency signal, extracting a respiration signal from the change of the difference frequency signal, and monitoring the respiration of the human body;
the step 4) performs signal processing on the difference frequency signal, and extracts a respiratory signal from the change of the difference frequency signal, which specifically includes:
4.1, calculating a PSD signal based on the difference frequency signal:
P＝k|S(f)| 2 ，
wherein,
in the formula, S (f) is a difference frequency signal, k is a scalar with a real value, wherein fs is a sampling frequency, w (i) is a window function, and L is the number of windows set when short-time Fourier transform is realized;
4.2, denoising the PSD signal, and eliminating the noise of the PSD signal caused by the self interference of the difference frequency signal through a self-adaptive signal normalization algorithm;
4.3, segmenting the PSD signal subjected to denoising processing, and inputting the segmented PSD signal into a slow feature analysis algorithm to extract a respiratory signal from an environment background;
4.4, detecting the number of peak values of respiratory waveforms in the respiratory signal by using a peak value monitoring algorithm, calculating the respiratory rate, and realizing the monitoring of human respiration;
the method for detecting the number of peaks of the respiratory waveform in the respiratory signal by using the peak monitoring algorithm and calculating the respiratory rate specifically comprises the following steps:
4.4.1, setting a sliding window length wn, a back window length wnb, and a breathing signal sequence B (t),
4.4.2, calculating the maximum value of the sliding window to be P, and the maximum value of the rear window to be S;
4.4.3, judging whether abnormal values appear in the two windows or not, if the abnormal values appear, eliminating the abnormal values, and if the abnormal values do not appear, continuing;
4.4.4, if the peak value of P is smaller than the peak value of S and the distance between the two points is smaller than a breathing interval, considering S as a breathing peak, and if the distance between the two points is larger than the breathing interval, considering P as a breathing peak, considering S as P, and performing the step 4.4.3; if the peak value of P is larger than S, the window moves backwards, and the step 4.4.3 is continued;
4.4.5, performing secondary scanning, and not in the threshold interval [ thr1, thr2 ]]Deleting the wave peak value; wherein, thr 1 ＝μ result -σ result ，thr 2 ＝μ result +σ result ，μ result Is the result mean, δ result Is the result standard deviation;
4.4.6, get the exact number of peaks n in the respiration signal.
2. The method for monitoring breathing on the basis of chirped continuous waves of inaudible sound according to claim 1, wherein the transmission signal is a high frequency chirp signal of 18-20 kHZ.
3. The method according to claim 1, wherein the step 2) of preprocessing the received signal and the transmitted signal to obtain the received signal and the transmitted signal after noise elimination comprises:
2.1, carrying out short-time Fourier transform on the received signal to obtain a time-frequency sequence of the received signal;
2.2, traversing the whole time-frequency sequence until a point with the first frequency not equal to 0 in the sequence is found and recording as an anchor point;
2.3, finishing traversing to obtain a preprocessed receiving signal;
2.4, preprocessing the transmitting signal, and obtaining a time-frequency sequence of the transmitting signal through short-time Fourier transform; traversing the whole time-frequency sequence until finding a point with the first frequency not equal to 0 in the sequence, and recording as an anchor point;
and 2.5, finishing traversing to obtain a preprocessed transmitting signal.
4. The method for monitoring respiration based on fm continuous waves of inaudible sounds according to claim 3, wherein the step 3) of mixing the pre-processed received signal with the transmission signal to obtain the difference frequency signal specifically comprises:
3.1, mixing the data between the transmission signal and the receiving signal after the preprocessing and the data between the signal from the anchor point to the signal end point to obtain a mixed signal;
3.2, selecting a Butterworth low-pass filter with the cutoff frequency of 1kHZ to eliminate high-frequency components in the mixed signal, and realizing low-pass filtering processing;
3.3, calculating the frequency difference between the transmitting signal and the receiving signal after the low-pass filtering processing to obtain a difference frequency signal; the difference frequency signal comprises a valid signal S E And an invalid signal S I Two parts, calculating the difference frequency signal f of the emission signal from the first sweep frequency period T through a formula b ：
Where n is the nth sweep period of the transmitted signal, n =1,2,3, \8230;, T is the sweep period, v is the speed of sound, and d is the distance between the device and the reflector.
5. The method of claim 1, wherein the specific procedure of removing the PSD signal noise caused by the self-interference of the difference frequency signal by the adaptive signal warping algorithm comprises:
(1) Setting a sliding window with the length of l to calculate the standard deviation of a difference frequency signal to obtain a standard deviation sequence s _ std = { s _ std1, s _ std2, \8230 \ 8230;, s _ stdn }, wherein the length of l is less than a frequency sweep period of a chirp signal, and the operation aims to reflect the volatility of an invalid signal to a PSD signal;
(2) Traversing the standard deviation sequence, adopting a dynamic threshold method to identify the starting point and the ending point of the invalid signal, and setting the length to be L 1 Calculating a mean value based on the sliding window signal as a dynamic threshold value
Realizing the identification and elimination of invalid signals;
(3) After the dynamic threshold is determined, the starting point and the end point of the invalid signal are identified, and the invalid signal is directly subtracted in the traversal process.
6. The method according to claim 1, wherein the noise-removed PSD signal is segmented and then input to a slow feature analysis algorithm to extract a respiratory signal from an environmental background, and the method specifically comprises:
4.3.1, dividing the input PSD signal into a plurality of equal-length segments with the lengths of n sampling points to finish first division;
4.3.2, re-slicing each equal-length fragment by using a time delay-based segmentation method, and incrementally increasing unit time delay for each equal-length fragment, wherein each re-slicing has an overlapping part;
4.3.3, finally, the PSD data N is divided into PSD data M; n, M is represented as:
wherein δ represents a delay of δ sample points;
and 4.3.4, inputting the matrix into a slow characteristic analysis algorithm as an input signal, and extracting a respiratory signal.
7. The method according to claim 6, wherein the matrix is input as an input signal to a slow feature analysis algorithm to extract the respiration signal, and the method specifically comprises:
inputting the data of the matrix into the slow characteristic SFA to obtain an output signal y (t), wherein the result in y (t) is according to delta (y) j (t)) arranging from small to large, and selecting a first row of y (t) as an optimal solution; if the optimal solution in y (t) can not reach the expected requirement, taking y (t) as the input function x (t), and repeating the steps until delta (y) is twice j (t)) a difference of less than 0.01;
wherein y (t) outputs a signal, the optimal result is that the jth component in y (t) is y j (t), then y j (t) satisfies: delta (y) j (t))＝＜y′ j (t)＞ 2 Minimum, y' j (t) is the first derivative of y (t).
8. The method of claim 1, wherein the calculated respiration rate is:
the peak value number n is the number of breaths, and the data length f is obtained by equal length division of the number of breaths n and the first time n Relation, calculating respiration rate v b ：
Wherein f is s Is the sampling frequency; v. of b Represents the breath rate in minutes in bpm, which represents the number of breaths per minute; according to the respiration rate v b And (4) calculating the number of breaths per minute, namely the breathing rate, and judging whether the abnormal breathing condition exists or not by using a formula.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110918251.8A CN113616188B (en) | 2021-08-11 | 2021-08-11 | Respiration monitoring method based on frequency modulated continuous waves of inaudible sound |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110918251.8A CN113616188B (en) | 2021-08-11 | 2021-08-11 | Respiration monitoring method based on frequency modulated continuous waves of inaudible sound |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113616188A CN113616188A (en) | 2021-11-09 |
| CN113616188B true CN113616188B (en) | 2022-10-04 |
Family
ID=78384385
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110918251.8A Active CN113616188B (en) | 2021-08-11 | 2021-08-11 | Respiration monitoring method based on frequency modulated continuous waves of inaudible sound |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113616188B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114469061B (en) * | 2022-01-20 | 2024-06-25 | 深圳市新起点健康服务有限公司 | Abdominal respiration monitoring system and method |
| CN114964013B (en) * | 2022-05-23 | 2024-07-23 | 天津大学四川创新研究院 | Terahertz film thickness measuring device and electronic equipment |
| CN115913392B (en) * | 2022-10-31 | 2023-06-23 | 中国矿业大学 | Potential pneumoconiosis early warning method based on ultrasonic signals |
| CN118161149B (en) * | 2024-05-14 | 2024-07-23 | 中国海洋大学 | Underwater breathing state monitoring method and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108113706A (en) * | 2017-12-19 | 2018-06-05 | 清华大学无锡应用技术研究院 | A kind of rhythm of the heart method, apparatus and system based on audio signal |
| CN109316174A (en) * | 2018-10-10 | 2019-02-12 | 大连理工大学 | A kind of vital sign wireless monitoring method based on Waveform Design and back wave processing |
| WO2021015559A1 (en) * | 2019-07-25 | 2021-01-28 | (주)비타 | Position tracking and respiration pattern measuring method and apparatus using fmcw radar |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058225A (en) * | 2019-04-09 | 2019-07-26 | 惠州市德赛西威智能交通技术研究院有限公司 | A kind of interior life entity detection system and method based on millimetre-wave radar |
| CN112674738A (en) * | 2020-12-07 | 2021-04-20 | 北京清雷科技有限公司 | Method and device for detecting respiration heartbeat signal |
-
2021
- 2021-08-11 CN CN202110918251.8A patent/CN113616188B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108113706A (en) * | 2017-12-19 | 2018-06-05 | 清华大学无锡应用技术研究院 | A kind of rhythm of the heart method, apparatus and system based on audio signal |
| CN109316174A (en) * | 2018-10-10 | 2019-02-12 | 大连理工大学 | A kind of vital sign wireless monitoring method based on Waveform Design and back wave processing |
| WO2021015559A1 (en) * | 2019-07-25 | 2021-01-28 | (주)비타 | Position tracking and respiration pattern measuring method and apparatus using fmcw radar |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113616188A (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113616188B (en) | Respiration monitoring method based on frequency modulated continuous waves of inaudible sound | |
| US11712198B2 (en) | Estimation of sleep quality parameters from whole night audio analysis | |
| US10765399B2 (en) | Programmable electronic stethoscope devices, algorithms, systems, and methods | |
| US11000223B2 (en) | Methods for detecting a sleep disorder and sleep disorder detection devices | |
| Schmidt et al. | Acoustic features for the identification of coronary artery disease | |
| CN103251388B (en) | Method and system of snoring monitoring and prevention and treatment based on smart phone platform | |
| CN102697520B (en) | Electronic stethoscope based on intelligent distinguishing function | |
| Allwood et al. | Advances in acoustic signal processing techniques for enhanced bowel sound analysis | |
| US20220007964A1 (en) | Apparatus and method for detection of breathing abnormalities | |
| JP6515670B2 (en) | Sleep depth estimation device, sleep depth estimation method, and program | |
| CN110037668B (en) | System for judging age, health state and malignant arrhythmia identification by combining pulse signal time-space domain with model | |
| US9591972B1 (en) | System and method for evaluating a state of health of a coronary artery | |
| CN105658142A (en) | Swallowing estimation device, information terminal device, and program | |
| US8419651B2 (en) | Spectrum analysis of coronary artery turbulent blood flow | |
| US20220225930A1 (en) | Apnea monitoring method and apparatus | |
| Schmidt et al. | Coronary artery disease and low frequency heart sound signatures | |
| KR20210032919A (en) | Method, system and non-transitory computer-readable recording medium for analyzing breath-related sounds | |
| Reddy et al. | Two microphones spectral-coherence based speech enhancement for hearing aids using smartphone as an assistive device | |
| Ghosh et al. | Heart sound data acquisition and preprocessing techniques: A review | |
| CN115015867B (en) | Identity identification and fall detection method based on ultra-wideband radar | |
| Saeidi et al. | 3D heart sound source localization via combinational subspace methods for long-term heart monitoring | |
| Schleusing et al. | Monitoring physiological and behavioral signals to detect mood changes of bipolar patients | |
| Sh-Hussain et al. | Classification of heart sound signals using autoregressive model and hidden markov model | |
| CN116115202A (en) | Method for detecting human breath and heartbeat frequency by ultra-wideband radar based on improved EMD | |
| Yang et al. | Lung water detection using acoustic techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |