CN110946584A - Method for recognizing human body sitting posture by using piezoresistive film pressure sensor group - Google Patents
Method for recognizing human body sitting posture by using piezoresistive film pressure sensor group Download PDFInfo
- Publication number
- CN110946584A CN110946584A CN201910992284.XA CN201910992284A CN110946584A CN 110946584 A CN110946584 A CN 110946584A CN 201910992284 A CN201910992284 A CN 201910992284A CN 110946584 A CN110946584 A CN 110946584A
- Authority
- CN
- China
- Prior art keywords
- sitting posture
- human body
- pressure
- vector machine
- support vector
- 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.)
- Pending
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1116—Determining posture transitions
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6887—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient mounted on external non-worn devices, e.g. non-medical devices
- A61B5/6892—Mats
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Physiology (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Power Engineering (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
A method for recognizing human sitting postures by utilizing piezoresistive film pressure sensor groups relates to a method for recognizing human sitting postures. A composite amplitude limiting filtering method is provided, which can filter abnormal values of original data in real time, so that the data has better smoothness and the fluctuation characteristics are not changed. Provides a human body sitting posture multi-classification algorithm of a two-layer structure support vector machine. The front-back axial inclination coefficient and the left-right axial inclination coefficient are defined to perform dimensionality reduction processing on pressure data, a support vector machine classifier is arranged on the first layer to perform classification calculation on normal sitting postures and abnormal sitting postures, two support vector machine classifiers are arranged on the second layer to perform classification calculation on human postures of the left front side, the right rear side, the right front side and the left rear side respectively in abnormal sitting postures, and human sitting posture recognition is completed through classification results and calculation.
Description
Technical Field
The invention relates to a human body sitting posture identification method, in particular to a human body sitting posture identification method by utilizing a piezoresistive film pressure sensor group.
Background
As early as 2003, a study by the world health organization indicated: every year, 200 million people worldwide die due to various diseases caused by sedentary life. Sedentariness has long been recognized as one of ten fatal, pathogenic killers. With the continuous acceleration of the rhythm of the modern society and the straight line increase of the living pressure, people need to face the problem of long-time sitting every day from reading to working of teenagers. And the sedentary people are bound to face the sitting posture problem. Teenagers need to sit for at least 10 hours every day from beginner, and most of their sitting postures are irregular, which greatly influences the growth and development of the teenagers. The sitting posture is a natural posture of the human body, and when the human body is in the sitting posture state, the energy consumption of the human body is reduced, and the fatigue is not easy to generate. When the human body is in different sitting postures, the pressure born by each part of the body is different. When a human body is in an incorrect sitting posture for a long time, the pressure born by buttocks, legs, the back, the spine and the like is very uneven, and great harm can be caused to the health of the human body in the past.
The harm to human health caused by poor sitting posture has three main aspects: causing cervical vertebra diseases of human body, aggravating symptoms of patients with lower back pain and causing myopia of teenagers. For office workers and student workers who sit for a long time, the harm to the physical health of the office workers and the student workers is caused by the poor sitting posture, and even the normal work and learning of the office workers and the student workers are affected in severe cases. When a human body is in a sitting posture, the left-leaning posture and the right-leaning posture can cause the deformation of the spine, and the serious people can cause the damage of the spine; the forward and backward leaning postures exacerbate the symptoms of partial lower back pain patients; if the sitting posture of the teenager is not standardized for a long time, the myopia of the eyes of the teenager is easily caused.
At present, the domestic similar technology is mainly based on expensive sensors such as an optical fiber sensor, a three-axis acceleration sensor and the like, and is mainly applied to high-end office environments. Although sitting posture monitoring is realized and an intelligent cushion capable of helping a user to improve sitting posture is developed, the intelligent cushion is not widely popularized in China because of high price.
Disclosure of Invention
The invention aims to provide a method for recognizing a human body sitting posture by utilizing a piezoresistive film pressure sensor group. The method not only can realize accurate recognition of the human body sitting posture, but also greatly reduces the hardware realization cost compared with the prior art.
The purpose of the invention is realized by the following technical scheme:
a method for identifying a human sitting posture by using a piezoresistive thin-film pressure sensor group, the method comprising the following steps:
(1) a piezoresistive pressure sensor arrangement mode based on a grouping strategy is designed:
arranging 8 or more piezoresistive film pressure sensor groups on a cushion for collecting pressure signals, and determining the arrangement mode of the sensor groups according to the thigh and hip pressure distribution characteristics and pressure measurement area grid division of a human body in a sitting posture state as the design basis of a grouping strategy; wherein, the No. 3, No. 4, No. 7 and No. 8 form four sensors to form two pressure measurement sensor groups for measuring the thigh pressure in a sitting posture state; the 1 st, the 2 nd, the 5 th and the 6 th form four sensors to form two pressure measurement sensor groups for collecting the pressure of the back hip part of the human body in a sitting posture state; the sensor grouping and arrangement mode acquires pressure transformation characteristics with obvious difference when a human body is in different sitting posture states;
(2) a composite amplitude limiting filtering method is provided:
preprocessing pressure data acquired in real time, filtering abnormal values and keeping data smoothness; the method comprises the following steps of adopting a box line graph method to judge abnormal values of pressure data and filter the abnormal values, and carrying out peak value limiting processing on the abnormal values and pressure amplitude values through a sliding window filtering method; smoothing the collected pressure data sequence by adopting a normal distribution weighted neighborhood mean value method; the pressure data after data preprocessing has fluctuation smoothness and does not change the fluctuation characteristics of the data;
(3) a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure is provided:
respectively defining front and rear axis inclination coefficients and left and right axis inclination coefficients, and performing dimensionality reduction processing on multi-dimensional pressure data according to the pressure sensor grouping; a support vector machine classifier is arranged on the first layer, two classification calculations of normal sitting posture and abnormal sitting posture are carried out according to the inclination coefficient, and classification is finished if the normal sitting posture is obtained; and if the sitting posture is not normal, starting the two support vector machine classifiers on the second layer to respectively perform two-class calculation on the human body postures of the left front part, the right rear part, the right front part and the left rear part so as to finish the human body sitting posture recognition.
The method for recognizing the human body sitting posture by utilizing the piezoresistive film pressure sensor group is characterized in that an abnormal value of pressure data is judged and filtered by adopting a box line graph method, and the abnormal value and a pressure amplitude value are subjected to peak value limiting processing by adopting a sliding window filtering method.
The method for recognizing the human body sitting posture by utilizing the piezoresistive thin-film pressure sensor group is characterized in that a composite amplitude limiting filtering method is to perform smoothing processing on a collected pressure data sequence by adopting a normal distribution weighted neighborhood mean value method.
The human body sitting posture identification method by utilizing the piezoresistive film pressure sensor group is characterized in that a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure respectively defines a front and back axis inclination coefficient and a left and right axis inclination coefficient according to pressure difference between pressure sensor groups and between the pressure sensor groups, and dimension reduction is carried out on preprocessed multidimensional pressure sensor data to two dimensions.
The method for recognizing the human body sitting posture by utilizing the piezoresistive film pressure sensor group is characterized in that a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure is that a support vector machine classifier is arranged on a first layer, and two classification calculations of a normal sitting posture and an abnormal sitting posture are carried out according to an axis inclination coefficient.
The human body sitting posture multi-classification algorithm of the support vector machine with the two-layer structure is a result of two-layer classification calculation of the support vector machine of the first layer, and classification is finished if the human body sitting posture multi-classification algorithm is normal; and if the sitting posture is not normal, starting the two support vector machine classifiers of the second layer.
The support vector machine human sitting posture multi-classification algorithm with the two-layer structure is two support vector machine classifiers on a second layer, carries out human posture two-classification calculation on the left front and the right back and the right front and the left back respectively, and finishes human sitting posture recognition through the AND operation of classification results.
The invention has the advantages and effects that:
1. according to the sensor grouping strategy and the sensor arrangement mode, pressure transformation characteristics with obvious differences can be acquired when a human body is in different sitting posture states, and the sensor grouping strategy and the sensor arrangement mode have good robustness.
2. The composite amplitude limiting filtering method provided by the invention can filter abnormal values of original data in real time, so that the data has better smoothness and the fluctuation characteristics are not changed.
3. The invention respectively defines the front and back axis inclination coefficients and the left and right axis inclination coefficients, reduces the dimension of the preprocessed multidimensional pressure sensor data to two dimensions, and has significant difference of the inclination axis coefficients under different human body sitting posture conditions.
4. The human body sitting posture multi-classification algorithm provided by the invention is a two-layer structure laminated support vector machine classifier, and can realize accurate classification of various human body sitting postures.
Drawings
FIG. 1 is an arrangement of piezoresistive thin film pressure sensor groups;
FIG. 2 is a graph of pressure sensor data pre-processed during experimental testing;
FIG. 3 is a state distribution scatter plot of the roll coefficient in experimental tests;
FIG. 4 is a flow chart of a human sitting posture multi-classification algorithm of a two-layer structure support vector machine;
FIG. 5 is a classification model of a first-tier support vector machine classifier in experimental testing;
FIG. 6 is a left-right-back-right-front classification model of a second-tier support vector machine classifier in experimental testing;
FIG. 7 is a front left, right, and back right classification model of the second tier support vector machine classifier in experimental testing.
In the figure: 1-8 are piezoresistive pressure sensors. Wherein, the four sensors of No. 3, No. 4, No. 7 and No. 8 form two pressure measurement sensor groups for measuring the thigh pressure in a sitting posture state; the four sensors 1, 2, 5 and 6 form two pressure measurement sensor groups for collecting the pressure of the back hip part of the human body in a sitting posture state. 9-12 are the four functional areas of the pressure foot. Where 9 and 10 are thigh pressure measuring regions, and 11 and 12 are hip pressure measuring regions.
Detailed Description
The present invention will be described in detail with reference to the embodiments shown in the drawings.
The invention relates to a human body sitting posture identification method based on a piezoresistive film pressure sensor group, which comprises three main parts: (1) the piezoresistive film pressure sensor group is adopted, the thigh and hip pressure distribution characteristics and the measurement area division of the human sitting posture state are used as the design basis of a grouping strategy and an arrangement mode, and the pressure characteristics with significant differences can be acquired in different sitting posture states. (2) A composite amplitude limiting filtering method is provided, which can filter abnormal values of original data in real time, so that the data has better smoothness and the fluctuation characteristics are not changed. (3) Provides a human body sitting posture multi-classification algorithm of a two-layer structure support vector machine. The front and back axial inclination coefficients and the left and right axial inclination coefficients are defined to reduce the dimension of pressure data to two dimensions, a support vector machine classifier is arranged on the first layer to perform classification calculation of normal sitting postures and abnormal sitting postures, two support vector machine classifiers are arranged on the second layer to perform classification calculation of human postures of the front left side and the rear right side and the front right side and the rear left side respectively for the abnormal sitting postures, and human sitting posture recognition is completed through classification results and calculation.
The invention provides a piezoresistive pressure sensor arrangement mode based on a grouping strategy. The method comprises the steps that piezoresistive film pressure sensor groups with the number of 8 or more are arranged on a cushion for collecting pressure signals, and the arrangement mode of the sensor groups is determined as the design basis of a grouping strategy according to the distribution characteristics of thigh and hip pressure and the grid division of a pressure measurement area when a human body is in a sitting posture state, as shown in the attached figure 1. Wherein, the four sensors of No. 3, No. 4, No. 7 and No. 8 form two pressure measurement sensor groups for measuring the thigh pressure in a sitting posture state; the four sensors 1, 2, 5 and 6 form two pressure measurement sensor groups for collecting the pressure of the back hip part of the human body in a sitting posture state.
In the invention, the grouping strategy and the arrangement mode of the piezoresistive pressure sensors are adopted. Experimental tests show that the sensor grouping strategy and the arrangement mode can acquire pressure transformation characteristics with obvious differences when a human body is in different sitting posture states, and the sensor grouping strategy and the arrangement mode have good robustness.
The invention provides a composite amplitude limiting filtering method. Preprocessing the pressure data acquired in real time, filtering abnormal values and keeping data smoothness. The abnormal value of the pressure data is judged and filtered by adopting a box curve graph method, and the peak value limiting processing is carried out on the abnormal value and the pressure amplitude value by adopting a sliding window filtering method. The abnormal value judging conditions of the box plot method are as follows: data represented by points outside the inner and outer limits of the boxplot are all abnormal values. In the sliding window filtering method, the window length is defined as Win _ L, the Mean value of data in the window is Win _ Mean, the peak value of data in the window is Win _ Max, and the suppression value of the peak value of data in the window is Win _ Dep, then the suppression value of the peak value of data in the window can be calculated as follows:
the invention provides a composite amplitude limiting filtering method. And smoothing the collected pressure data sequence by adopting a normal distribution weighted neighborhood mean value method. The pressure data after data preprocessing has better fluctuation smoothness and does not change the fluctuation characteristics of the data. Defining the current Data as Data _ Cur, the Left neighborhood Data as Data _ Left, the Right neighborhood Data as Data _ Right, and the smoothed Data as Data _ Smt. The standard normal distribution weighting coefficient takes the values of the function of x =0 and x =1 in the normal distribution function table, and is 0.5 and 0.84 respectively. Then Data _ Smt is calculated as follows:
the invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. According to the pressure difference between the pressure sensor groups and the pressure difference between the pressure sensor groups, front and back axis inclination coefficients and left and right axis inclination coefficients are respectively defined, dimensionality reduction is carried out on the preprocessed multidimensional pressure sensor data to two dimensions, and the axis inclination coefficients have significant differences under different human body sitting posture conditions. The multidimensional sensor data dimension reduction calculation process is as follows:
the pressure monitoring data of the sensors 1 to 8 shown in fig. 1 are defined as DataS1 to DataS8, respectively. The group of sensors including sensor 1 and sensor 2 has a mean value of pressure monitoring data defined as DataG1, the group of sensors including sensor 5 and sensor 6 has a mean value of pressure monitoring data defined as DataG2, the group of sensors including sensor 3 and sensor 4 has a mean value of pressure monitoring data defined as DataG3, and the group of sensors including sensor 7 and sensor 8 has a mean value of pressure monitoring data defined as DataG 4. The anterior-posterior and left-right roll coefficients are defined as AIC _ FR and AIC _ LR, respectively, and there are:
the invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. The first layer is provided with a support vector machine classifier, and the classification calculation of the normal sitting posture and the abnormal sitting posture of the human body is carried out according to the inclination coefficient. The algorithm flow and the two-layer structure support the classifier set of the vector machine, as shown in fig. 4. The human normal and abnormal sitting posture classification model is shown in figure 5.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And if the classification result of the first-layer support vector machine classifier is that the human body is in a normal sitting posture, finishing the classification. The human normal and abnormal sitting posture classification model is shown in figure 5.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And the second layer is provided with two support vector machine classifiers for carrying out classification calculation on the human body abnormal sitting postures of the left front, the right back, the right front and the left back respectively. Left front and right rear sitting posture classification models, as shown in figure 6. The right front and left rear sitting posture classification models are shown in fig. 7.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And if the classification result of the first layer of support vector machine classifier is that the human body is not in a normal sitting posture, starting the two support vector machine classifiers of the second layer.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And according to the two classification results of the human body abnormal sitting postures of the left front, the right back, the right front and the left back obtained by the two support vector machine classifiers in the second layer through calculation, the human body sitting posture recognition is completed through the mutual AND operation of the classification results.
Example 1
The invention provides a piezoresistive pressure sensor arrangement mode based on a grouping strategy. 8 piezoresistive pressure sensors are used. Wherein, the four sensors of No. 3, No. 4, No. 7 and No. 8 form two pressure measurement sensor groups for measuring the thigh pressure in a sitting posture state; the four sensors 1, 2, 5 and 6 form two pressure measurement sensor groups for collecting the pressure of the back hip part of the human body in a sitting posture state. 9-12 are the four functional areas of the pressure foot. Where 9 and 10 are thigh pressure measuring regions, and 11 and 12 are hip pressure measuring regions.
According to the piezoresistive pressure sensor grouping strategy and the piezoresistive pressure sensor arrangement method, experimental tests show that the sensor grouping strategy and the sensor arrangement method can acquire pressure transformation characteristics with obvious differences when a human body is in different sitting posture states, and the piezoresistive pressure sensor grouping strategy and the piezoresistive pressure sensor arrangement method have good robustness. The pressure monitoring data is a pre-processed profile as shown in fig. 2. The scatter diagram of the axial inclination coefficient distribution of different human sitting postures processed to two dimensions by the dimension reduction is shown in fig. 3.
The invention provides a composite amplitude limiting filtering method. Preprocessing the pressure data acquired in real time, filtering abnormal values and keeping data smoothness. The abnormal value of the pressure data is judged and filtered by adopting a box curve graph method, and the peak value limiting processing is carried out on the abnormal value and the pressure amplitude value by adopting a sliding window filtering method. The abnormal value judging conditions of the box plot method are as follows: data represented by points outside the inner and outer limits of the boxplot are all abnormal values. In the sliding window filtering method, the window length is defined as Win _ L, the Mean value of data in the window is Win _ Mean, the peak value of data in the window is Win _ Max, and the suppression value of the peak value of data in the window is Win _ Dep, then the suppression value of the peak value of data in the window can be calculated as follows:
the invention provides a composite amplitude limiting filtering method. And smoothing the collected pressure data sequence by adopting a normal distribution weighted neighborhood mean value method. The pressure data after data preprocessing has better fluctuation smoothness and does not change the fluctuation characteristics of the data. Defining the current Data as Data _ Cur, the Left neighborhood Data as Data _ Left, the Right neighborhood Data as Data _ Right, and the smoothed Data as Data _ Smt. The standard normal distribution weighting coefficient takes the values of the function of x =0 and x =1 in the normal distribution function table, and is 0.5 and 0.84 respectively. Then Data _ Smt is calculated as follows:
in the invention, the provided composite amplitude limiting filtering method shows that the pressure data after data preprocessing has better fluctuation smoothness and does not change the data fluctuation characteristics in the test. The data distribution graph after the pressure monitoring data preprocessing is shown in fig. 2.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. According to the pressure difference between the pressure sensor groups and the pressure difference between the pressure sensor groups, front and back axis inclination coefficients and left and right axis inclination coefficients are respectively defined, dimensionality reduction is carried out on the preprocessed multidimensional pressure sensor data to two dimensions, and the axis inclination coefficients have significant differences under different human body sitting posture conditions. The multidimensional sensor data dimension reduction calculation process is as follows:
as shown in fig. 1, the pressure monitoring data of the sensors 1 to 8 are defined as data 1 to data 8, respectively. The group of sensors including sensor 1 and sensor 2 has a mean value of pressure monitoring data defined as DataG1, the group of sensors including sensor 5 and sensor 6 has a mean value of pressure monitoring data defined as DataG2, the group of sensors including sensor 3 and sensor 4 has a mean value of pressure monitoring data defined as DataG3, and the group of sensors including sensor 7 and sensor 8 has a mean value of pressure monitoring data defined as DataG 4. The anterior-posterior and left-right roll coefficients are defined as AIC _ FR and AIC _ LR, respectively, and there are:
the invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. According to the pressure sensor grouping, front and back axis inclination coefficients and left and right axis inclination coefficients are respectively defined, dimensionality reduction is carried out on the preprocessed multidimensional pressure sensor data to be in two dimensions, and the axis inclination coefficients have significant differences under different human body sitting posture conditions. The multidimensional data of the pressure sensor is subjected to dimension reduction processing to obtain two-dimensional distribution scatter diagrams of the axil tilt coefficients of different human body sitting postures, as shown in fig. 3.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. The first layer is provided with a support vector machine classifier, and the classification calculation of the normal sitting posture and the abnormal sitting posture of the human body is carried out according to the inclination coefficient. The algorithm flow and the two-layer structure support the classifier set of the vector machine, as shown in fig. 4. The human normal and abnormal sitting posture classification model is shown in figure 5.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And if the classification result of the first-layer support vector machine classifier is that the human body is in a normal sitting posture, finishing the classification. The algorithm flow and the two-layer structure support the classifier set of the vector machine, as shown in fig. 4.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And the second layer is provided with two support vector machine classifiers for carrying out classification calculation on the human body abnormal sitting postures of the left front, the right back, the right front and the left back respectively. The algorithm flow and the two-layer structure support the classifier set of the vector machine, as shown in fig. 4. Left front and right rear sitting posture classification models, as shown in figure 6. The right front and left rear sitting posture classification models are shown in fig. 7.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And if the classification result of the first layer of support vector machine classifier is that the human body is not in a normal sitting posture, starting the two support vector machine classifiers of the second layer. The algorithm flow and the two-layer structure support the classifier set of the vector machine, as shown in fig. 4.
The invention provides a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure. And according to the two classification results of the human body abnormal sitting postures of the left front, the right back, the right front and the left back obtained by the two support vector machine classifiers in the second layer through calculation, the human body sitting posture recognition is completed through the mutual AND operation of the classification results. The algorithm flow and the two-layer structure support the classifier set of the vector machine, as shown in fig. 4.
Claims (7)
1. A method for recognizing a human body sitting posture by utilizing a piezoresistive film pressure sensor group is characterized by comprising the following steps:
(1) a piezoresistive pressure sensor arrangement mode based on a grouping strategy is designed:
arranging 8 or more piezoresistive film pressure sensor groups on a cushion for collecting pressure signals, and determining the arrangement mode of the sensor groups according to the thigh and hip pressure distribution characteristics and pressure measurement area grid division of a human body in a sitting posture state as the design basis of a grouping strategy; wherein, the No. 3, No. 4, No. 7 and No. 8 form four sensors to form two pressure measurement sensor groups for measuring the thigh pressure in a sitting posture state; the 1 st, the 2 nd, the 5 th and the 6 th form four sensors to form two pressure measurement sensor groups for collecting the pressure of the back hip part of the human body in a sitting posture state; the sensor grouping and arrangement mode acquires pressure transformation characteristics with obvious difference when a human body is in different sitting posture states;
(2) a composite amplitude limiting filtering method is provided:
preprocessing pressure data acquired in real time, filtering abnormal values and keeping data smoothness; the method comprises the following steps of adopting a box line graph method to judge abnormal values of pressure data and filter the abnormal values, and carrying out peak value limiting processing on the abnormal values and pressure amplitude values through a sliding window filtering method; smoothing the collected pressure data sequence by adopting a normal distribution weighted neighborhood mean value method; the pressure data after data preprocessing has fluctuation smoothness and does not change the fluctuation characteristics of the data;
(3) a support vector machine human body sitting posture multi-classification algorithm with a two-layer structure is provided:
respectively defining front and rear axis inclination coefficients and left and right axis inclination coefficients, and performing dimensionality reduction processing on multi-dimensional pressure data according to the pressure sensor grouping; a support vector machine classifier is arranged on the first layer, two classification calculations of normal sitting posture and abnormal sitting posture are carried out according to the inclination coefficient, and classification is finished if the normal sitting posture is obtained; and if the sitting posture is not normal, starting the two support vector machine classifiers on the second layer to respectively perform two-class calculation on the human body postures of the left front part, the right rear part, the right front part and the left rear part so as to finish the human body sitting posture recognition.
2. The method for recognizing the sitting posture of the human body by using the piezoresistive film pressure sensor group as claimed in claim 1, wherein the composite amplitude limiting filtering method is to use a boxplot method to discriminate the abnormal value of the pressure data and filter the abnormal value, and to use a sliding window filtering method to perform peak limiting processing on the abnormal value and the pressure amplitude.
3. The method for recognizing the sitting posture of the human body by using the piezoresistive thin-film pressure sensor group as claimed in claim 1, wherein the composite amplitude limiting filtering method is to smooth the collected pressure data sequence by adopting a normal distribution weighted neighborhood mean value method.
4. The method for recognizing the human sitting posture by utilizing the piezoresistive thin-film pressure sensor group according to claim 1, wherein the support vector machine human sitting posture multi-classification algorithm with the two-layer structure is used for respectively defining a front-back axis inclination coefficient and a left-right axis inclination coefficient according to the pressure difference between the groups and the groups of the pressure sensors, and reducing the dimension of the preprocessed multi-dimensional pressure sensor data to two dimensions.
5. The method for human body sitting posture identification by utilizing the piezoresistive film pressure sensor group according to claim 4, wherein the support vector machine human body sitting posture multi-classification algorithm with the two-layer structure is characterized in that a support vector machine classifier is arranged on a first layer, and two classification calculations of a normal sitting posture and an abnormal sitting posture are carried out according to an inclination coefficient.
6. The method for recognizing the human sitting posture by using the piezoresistive film pressure sensor group according to claim 5, wherein the support vector machine human sitting posture multi-classification algorithm with the two-layer structure is the result of the classification calculation of the support vector machine of the first layer, and the classification is finished if the human sitting posture is normal; and if the sitting posture is not normal, starting the two support vector machine classifiers of the second layer.
7. The method for recognizing the sitting posture of the human body by using the piezoresistive film pressure sensor group as claimed in claim 6, wherein the support vector machine multi-classification algorithm of the two-layer structure is two support vector machine classifiers at the second layer, which perform two-classification calculation of the human body postures of the front left and the back right and the front right and the back left respectively, and the human body sitting posture recognition is completed through the AND operation of the classification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910992284.XA CN110946584A (en) | 2019-10-18 | 2019-10-18 | Method for recognizing human body sitting posture by using piezoresistive film pressure sensor group |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910992284.XA CN110946584A (en) | 2019-10-18 | 2019-10-18 | Method for recognizing human body sitting posture by using piezoresistive film pressure sensor group |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110946584A true CN110946584A (en) | 2020-04-03 |
Family
ID=69975596
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910992284.XA Pending CN110946584A (en) | 2019-10-18 | 2019-10-18 | Method for recognizing human body sitting posture by using piezoresistive film pressure sensor group |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110946584A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111345633A (en) * | 2020-04-08 | 2020-06-30 | 时新 | Intelligent induction pad |
| CN112949434A (en) * | 2021-02-19 | 2021-06-11 | 清华大学 | Human body sitting posture recognition system based on pressure cushion |
| CN113317937A (en) * | 2021-05-28 | 2021-08-31 | 金陵科技学院 | Internet of things-based home-based auxiliary intelligent wheelchair for old people and control method |
| CN113361342A (en) * | 2021-05-20 | 2021-09-07 | 杭州麦淘淘科技有限公司 | Multi-mode-based human body sitting posture detection method and device |
| CN113627236A (en) * | 2021-06-24 | 2021-11-09 | 广东技术师范大学 | Sitting posture identification method, device, equipment and storage medium |
| CN114419676A (en) * | 2022-01-24 | 2022-04-29 | 平安国际智慧城市科技股份有限公司 | Sitting posture analysis method and device based on artificial intelligence, computer equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH10318861A (en) * | 1997-05-21 | 1998-12-04 | Furukawa Electric Co Ltd:The | Seating detecting device |
| CN107554527A (en) * | 2017-08-04 | 2018-01-09 | 武汉理工大学 | One kind has driving condition supervision and pre-warning system and method |
-
2019
- 2019-10-18 CN CN201910992284.XA patent/CN110946584A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH10318861A (en) * | 1997-05-21 | 1998-12-04 | Furukawa Electric Co Ltd:The | Seating detecting device |
| CN107554527A (en) * | 2017-08-04 | 2018-01-09 | 武汉理工大学 | One kind has driving condition supervision and pre-warning system and method |
Non-Patent Citations (4)
| Title |
|---|
| 余胜威等: "《MATLAB图像滤波去噪分析及其应用》", 30 September 2015 * |
| 刘少华: "基于柔性阵列压力传感器的坐姿监测与提醒系统的设计与实现", 《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》 * |
| 花蕾: "预测算法在广告效应中的应用研究", 《中国优秀博硕士学位论文全文数据库（硕士） 经济与管理科学辑》 * |
| 谢中华等: "《新编MATLAB/Simulink自学一本通》", 31 January 2018 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111345633A (en) * | 2020-04-08 | 2020-06-30 | 时新 | Intelligent induction pad |
| CN112949434A (en) * | 2021-02-19 | 2021-06-11 | 清华大学 | Human body sitting posture recognition system based on pressure cushion |
| CN113361342A (en) * | 2021-05-20 | 2021-09-07 | 杭州麦淘淘科技有限公司 | Multi-mode-based human body sitting posture detection method and device |
| CN113317937A (en) * | 2021-05-28 | 2021-08-31 | 金陵科技学院 | Internet of things-based home-based auxiliary intelligent wheelchair for old people and control method |
| CN113627236A (en) * | 2021-06-24 | 2021-11-09 | 广东技术师范大学 | Sitting posture identification method, device, equipment and storage medium |
| CN114419676A (en) * | 2022-01-24 | 2022-04-29 | 平安国际智慧城市科技股份有限公司 | Sitting posture analysis method and device based on artificial intelligence, computer equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110946584A (en) | Method for recognizing human body sitting posture by using piezoresistive film pressure sensor group | |
| US20210196055A1 (en) | Bedding System with a CNN Based Machine Vision Process | |
| CN104732250B (en) | The sleeping position and sleep behavior of a kind of noiseless sleep test recognition methods | |
| CN109820373A (en) | Sitting posture self-adapting regulation method based on Intelligent seat | |
| CN101630410A (en) | Human body sitting posture judgment method based on single camera | |
| CN110532850B (en) | Fall detection method based on video joint points and hybrid classifier | |
| WO2020019358A1 (en) | Physiological parameter measuring system and intelligent seat having same | |
| CN109171315A (en) | Automatically adjust mattress and intelligent bed | |
| CN112472048B (en) | Method for realizing neural network for identifying pulse condition of cardiovascular disease patient | |
| CN106056116A (en) | Fuzzy rough set-based sleeping posture pressure image recognition method | |
| CN110197235B (en) | Human body activity recognition method based on unique attention mechanism | |
| CN113269096B (en) | Head pressure data deep learning processing method for sleeping gesture recognition | |
| CN107169453A (en) | A kind of sitting posture detecting method based on depth transducer | |
| CN109091150A (en) | Recognition methods, sleep quality appraisal procedure and the intelligent wearable device that body of sleeping moves | |
| CN103838963A (en) | Bra pressure comfort evaluation method | |
| CN109512188A (en) | A kind of sitting posture detecting method and device and seat | |
| CN112949434A (en) | Human body sitting posture recognition system based on pressure cushion | |
| CN113688720A (en) | Neural network recognition-based sleeping posture prediction method | |
| CN110063736A (en) | The awake system of fatigue detecting and rush of eye movement parameter monitoring based on MOD-Net network | |
| CN111104740B (en) | Clothing pressure measurement method and system based on virtual human model | |
| CN106820763A (en) | Mattress comfortableness system of selection based on pressure distribution | |
| CN109674477A (en) | Computer vision Postural Analysis method based on deep learning | |
| CN116097198A (en) | Determining characteristics from biometric signals | |
| CN113128031A (en) | Measuring device and evaluation method for comfort level of seat | |
| CN115127704B (en) | Sitting posture identification method based on flexible film pressure sensor |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200403 |