Nothing Special   »   [go: up one dir, main page]

WO2024093188A1 - 信号检测方法、信号处理方法、信号处理模型 - Google Patents

信号检测方法、信号处理方法、信号处理模型 Download PDF

Info

Publication number
WO2024093188A1
WO2024093188A1 PCT/CN2023/093732 CN2023093732W WO2024093188A1 WO 2024093188 A1 WO2024093188 A1 WO 2024093188A1 CN 2023093732 W CN2023093732 W CN 2023093732W WO 2024093188 A1 WO2024093188 A1 WO 2024093188A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
value
point
segment
mapping
Prior art date
Application number
PCT/CN2023/093732
Other languages
English (en)
French (fr)
Inventor
黄肖山
胥红来
章希睿
宫长辉
梁星
Original Assignee
博睿康科技（常州）股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 博睿康科技（常州）股份有限公司 filed Critical 博睿康科技（常州）股份有限公司
Publication of WO2024093188A1 publication Critical patent/WO2024093188A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques

Definitions

  • the present invention relates to the field of signal detection technology, and in particular to a signal detection method, a signal processing method, and a signal processing model.
  • the collected analog signals have DC bias, which greatly increases the dynamic range of the original data, and thus requires a larger bit-width ADC to complete high-fidelity analog-to-digital conversion.
  • the existing signal transmission and storage methods do not detect or process the signals in stages, resulting in a larger ADC bit width, higher requirements for signal transmission and storage, and an inevitable increase in hardware costs.
  • the technical problem to be solved by the present invention is to provide a signal detection method, a signal processing method and a signal processing model.
  • the signal detection method includes: signal preprocessing, that is, processing the signal points in any signal segment to obtain the processing value of the signal point; detecting whether the processing value exceeds the set overflow threshold range, so as to select different reprocessing methods according to the detection result of the processing value.
  • the processing value is a filtered processing value or a residual value of a signal point measured by a signal preprocessing model
  • the signal preprocessing model includes: any one of an AR model, a single-point difference model, a periodic difference model, and a combination model of a single-point difference and a periodic difference; when the signal has no periodic interference and the sampling rate is high, the single-point difference model is used to measure the residual value of the signal point; when the signal has periodic interference, the periodic difference model is used to measure the residual value of the signal point; when the signal has periodic interference and the sampling rate is high, the combination model of a single-point difference and a periodic difference is used to measure the residual value of the signal point.
  • the signal detection method also includes performing phased detection on the signal segments before signal preprocessing; the phased detection includes: sliding window processing; detecting the volatility of the signal within the window; and detecting the signal phases according to the volatility results of the signal within the window.
  • the present invention also provides a signal processing method based on the signal detection method as described above, wherein the different reprocessing methods are selected according to the detection results of the processing values, including: when the processing value of any signal point exceeds the set overflow threshold range, the processing value of the signal point is mapped and encoded; wherein the mapping encoding includes: repeatedly performing N mapping operations on the processing value of the signal point until it is within the overflow threshold range, and obtaining a mapping mark and a mapping value, then the signal point is encoded as: N mapping marks + M mapping values; wherein M is the type of mapping value, M ⁇ 1.
  • mapping operation includes a subtraction operation; the subtraction operation is to subtract the overflow threshold from the processed value of the signal point; the mapping value is the difference obtained by the last subtraction operation, then the signal point is encoded as: N mapping marks + difference.
  • the overflow threshold range is determined by an overflow threshold, the overflow threshold is -2n -1 or 2n-1 , n is the expected coding bit width, and the overflow threshold range is (-2n -1 , 2n -1 ); the n-bit overflow rate of the signal segment is set equal to the number of signal points in any signal segment whose processing values exceed the overflow threshold range divided by the total number of signal points in the signal segment; when the signal segment is in a stable period and the n-bit overflow rate of the signal segment is lower than a first threshold, a mapping code using a subtraction operation is selected to encode the processing value; when the signal segment is in a fluctuating period and the average information entropy of the signal segment is less than a set entropy threshold, a mapping code using a logarithmic operation is selected to encode the processing value; when the signal segment is in a fluctuating period and the average information entropy of the signal segment is not less than a set entropy threshold or the signal segment is in a
  • the selecting of different reprocessing methods according to the detection result of the processing value also includes:
  • the signal points or the signal points whose processing values do not exceed the set overflow threshold range are secondary encoded;
  • the secondary encoding includes: Huffman coding and its variant coding, arithmetic coding and its variant coding, interval coding and its variant coding, and at least one of asymmetric digital system coding.
  • the present invention also provides a signal processing model, comprising: an acquisition module for acquiring signals; a processing module for executing the signal processing method as described above; and a storage module for storing processed signals.
  • the beneficial effect of the present invention is that the signal processing method and signal processing model of the present invention first perform signal preprocessing on the signal to obtain the processing value of the signal point; then detect whether the processing value exceeds the set overflow threshold range; and then select different reprocessing methods according to the detection result of the processing value, effectively making targeted differentiation and processing on the signal according to the detection result of the processing value, thereby effectively improving the processing effect of the signal.
  • FIG1 is a flow chart of a signal processing method of the present invention.
  • FIG. 2 is a diagram showing a specific working process of the signal processing method of the present invention.
  • FIG. 3 is a schematic diagram of the quotient-remainder operation coding of the present invention.
  • FIG. 4 is a schematic diagram of periodic differential and single-point differential operations of the present invention.
  • FIG. 5 is a comparison diagram of overflow rates of different signal preprocessing models of the present invention.
  • FIG. 6 is another comparison diagram of the overflow rates of different signal preprocessing models of the present invention.
  • FIG. 7 is a processing diagram of the first embodiment of the present invention.
  • FIG8 is a processing diagram of the second embodiment of the present invention.
  • FIG. 9 is a processing diagram of the third embodiment of the present invention.
  • FIG. 10 is a process diagram of a comparative example.
  • FIG. 11 is a graph showing changes in compression ratios of the embodiments of the present invention and the comparative example.
  • FIG. 12 is a graph showing changes in the overflow rate of the embodiment of the present invention and the comparative example.
  • the terms “installed”, “connected”, and “connected” should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components.
  • installed should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components.
  • the signal processing method of the present invention includes the following steps: first, a signal detection method is used to detect whether the processing value of the signal exceeds the set overflow threshold range, and then different reprocessing methods are selected according to the detection result of the processing value.
  • the signal detection method includes: S1, signal preprocessing, that is, processing the signal point in any signal segment to obtain the processing value of the signal point; S2, detecting whether the processing value exceeds the set overflow threshold range; and selecting different reprocessing methods according to the detection result of the processing value.
  • Selecting different reprocessing methods according to the detection result of the processing value includes: when the processing value of any signal point exceeds the set overflow threshold range, first mapping and encoding the processing value of the signal point, and then performing secondary encoding; when the processing value of any signal point does not exceed the set overflow threshold range, directly performing secondary encoding on its processing value; wherein the mapping encoding includes: repeatedly performing N mapping operations on the processing value of the signal point until it is within the overflow threshold range, obtaining a mapping mark and a mapping value, then the signal point is encoded as: N mapping marks + M types of mapping values; wherein M is the type of mapping value, M ⁇ 1.
  • the M-bit original signal can obtain multiple M-bit signal segments after sliding window processing.
  • the signal preprocessing model is used to calculate the processing value of the signal point, and then it is determined whether the processing value of the signal point exceeds the overflow threshold range. If so, mapping coding processing is performed, if not, the processing value is saved, and then the mapped coded processing value and the directly saved processing value are both re-encoded to obtain the final signal.
  • the secondary coding includes but is not limited to: at least one of Huffman coding and its variant coding, arithmetic coding and its variant coding, interval coding and its variant coding, and asymmetric digital system coding.
  • the signal after signal encoding can be further secondary encoded to further compress the signal.
  • the signal mentioned in the present invention can be any type of signal, such as but not limited to physiological signals such as EEG signals, EMG signals, and ECG signals, and can also be storable data such as image signals and software data.
  • the original signal can be sliced and divided into multiple signal segments, and then the signal segments are detected in stages, and the signal stages are marked according to the volatility, and then the signal points in each signal segment are preprocessed and encoded to obtain to a compressed signal.
  • the processing value is the residual value or filtered processing value of the signal point measured by the signal preprocessing model. The processing values of all signal points in the signal segment are judged.
  • the encoding method of the signal point is "N mapping marks + M mapping values", where the mapping mark is to facilitate the subsequent decoding of the compressed signal, and the mapping value is at least one.
  • the mapping operation includes a subtraction operation.
  • the subtraction operation is to subtract the overflow threshold from the processed value of the signal point, and the mapping value is the difference obtained by the last subtraction operation.
  • the signal point is encoded as: N mapping marks + difference.
  • the mapping value is the subtraction difference.
  • the overflow threshold is subtracted from the processed value of the signal point to obtain the difference of the first subtraction. If the difference of the first subtraction still exceeds the overflow threshold range, the subtraction operation continues to be performed on the difference until the difference is within the overflow threshold range.
  • the mapping mark can be a character
  • N represents the number of subtractions
  • the mapping value selects the difference obtained by the last subtraction.
  • the mapping operation includes any one of logarithmic operation and quotient-remainder operation.
  • D is the processed value
  • the mapping value is the floating-point value Z obtained by the last logarithmic operation, whose integer segment is int(Z) and the decimal segment is Z-int(Z); then the signal point is encoded as: N mapping marks + int(Z) + Z-int(Z).
  • the quotient-remainder operation is to divide the processed value of the signal point by the overflow threshold, and the mapping value is the quotient value Q obtained by the last quotient-remainder operation and the N remainders C obtained by N quotient-remainder operations, then the signal point is encoded as: N mapping marks + quotient value Q + N remainders C. That is to say, when performing a logarithmic operation, if the floating-point value obtained after the logarithmic operation on the processed value of the signal point still exceeds the overflow threshold range, then the logarithmic operation on the floating-point value continues until the floating-point value is within the overflow threshold range.
  • the mapping mark can be a character, N represents the number of times the logarithmic operation is performed, and the mapping value includes two types: integer segment int(Z) and decimal segment Z-int(Z).
  • the integer segment int(Z) only occupies 1 byte, and the decimal segment Z-int(Z) can occupy multiple bytes.
  • the mapping mark can be a character
  • N represents the number of times the quotient and remainder operation is performed
  • the mapping value includes two types: quotient Q and remainder C.
  • the quotient Q selects the quotient obtained by the last division, and the remainder C retains all.
  • the quotient and remainder operation maps the signal point to a smaller quotient and remainder by performing division.
  • the encoding method of the present invention can compress the data volume of the signal as much as possible while retaining the signal characteristics, thereby further improving the compression ratio.
  • the overflow threshold range is determined by the overflow threshold, the overflow threshold is -2n -1 or 2n-1 , n is the expected encoding bit width, then the overflow threshold range is (-2n -1 , 2n -1 ), and the overflow threshold range does not include the endpoints.
  • the overflow threshold participates in the mapping operation, its value depends on whether the number with the largest absolute value in the numerical value represented by the computer is positive or negative. In other words, its value is It depends on the signal range that the computer's n-bit signed signal can represent.
  • the overflow threshold is 2n -1 ; if it is [-2n -1 , 2n -1 ), the overflow threshold is -2n -1 .
  • the same computer performs signal compression, it should be unified to one of the positive or negative numbers.
  • the first threshold is 1-10%, preferably 3%, 5%, 7%. That is to say, when selecting the signal coding method, the coding method to be selected can be determined according to the n-bit overflow rate of the signal point in the signal segment, the signal fluctuation state (stable period or fluctuating period) and the average information entropy of the signal segment. Choose a suitable encoding method based on the characteristics of the signal. For example, if the overflow rate of a signal in a stable period is small, there is no need to use logarithmic operations and quotient operations with relatively large computational complexity. This can not only better compress the signal, save measurement resources, and improve efficiency, but also retain important features in the signal.
  • the overflow threshold is -128 or 128, the overflow threshold range is (-128, 128), and the original signal point is 24 bits, and the signal segment contains 10 signal points.
  • Single-point difference is performed on the signal points in the signal segment to obtain 9 residual values, and the overflow rate is 50%. Therefore, the quotient and remainder operation can be selected to encode the 10 signal points (1 original value + 9 residual values) in the processed signal segment.
  • the overflow threshold can be set to 128.
  • the encoding of all signal points in the signal segment can be obtained through the quotient-remainder operation (as shown in Figure 3). After signal encoding, the 10 signals are reduced from 30 bytes to 22 bytes, which effectively reduces the signal volume and saves storage space.
  • the quotient remainder operation is a lossless compression, and the number of bits of the encoded signal is equal to the number of bits of the original signal, so that after decoding, a valid signal (ie, the original characteristics are retained) can still be obtained for subsequent analysis and processing.
  • the signal detection method further includes performing phased detection on the signal segments before signal preprocessing; the phased detection includes: sliding window processing; detecting the volatility of the signal within the window; and detecting the signal phases according to the volatility results of the signal within the window. That is, before performing signal preprocessing, the signal segments can be classified to determine whether the signal segments are in a stable period or a fluctuating period, and then the signal preprocessing and selective reprocessing can be performed according to the classification results of the signal segments. In this way, To further improve the signal processing effect, that is, the compression effect and compression efficiency after signal processing.
  • the volatility of the signal in the window is characterized by any one of the variance and mean of the signal in the window and the normalized line length of the signal in the window; wherein detecting the signal phase according to the variance and mean of the signal in the window comprises: obtaining the variance and mean of the signal in the window; detecting a signal segment whose variance is not less than a second threshold as a fluctuation period, and detecting a signal segment whose variance is less than the second threshold as a stable period.
  • the second threshold is 2-10 times the mean of the variance of the signal in the window, preferably 3 times, 5 times, or 8 times.
  • the processing method of the filtering value mainly adopts low-pass filtering or high-pass filtering to eliminate background noise unrelated to the target signal.
  • the signal preprocessing model mainly adopted includes: any one of an AR model, a single-point difference model, a periodic difference model, and a combination model of a single-point difference and a periodic difference.
  • the single-point difference model is used to measure the residual value of the signal point; when the signal has periodic interference, the periodic difference model is used to measure the residual value of the signal point; when the signal has periodic interference and the sampling rate is high, the single-point difference and the periodic difference combination model is used to measure the residual value of the signal point.
  • the expression of the AR model is:
  • t P + 1, ..., K, is the coefficient of the P-order centralized AR model, is the residual between the original signal and its estimated value.
  • ⁇ [t] is the key to improving the compression ratio.
  • the residual vector of the signal segment is:
  • the coefficients of the minimized AR model can be expressed as:
  • the single-point difference model can reduce the overflow rate of the residual value.
  • a periodic difference model is obtained, where T represents the maximum period of interference, Fs is the sampling rate, and w is the trend compensation coefficient, which can reflect the rate of change of the overall replication of the previous and next periods.
  • the periodic difference model can remove the periodic interference in the original signal, thereby reducing the overflow rate of the residual value.
  • the signal preprocessing model can be selected according to whether the signal has periodic interference and the sampling rate.
  • T 1/50 or 1/60 (maximum period of power frequency interference), and the number of signal points corresponding to the maximum interference period is T*Fs.
  • a certain signal segment contains signal points y 1 , y 2 , ... , y K , a total of K signal points, and divide the K signal points into multiple periodic segments y 1 , y 2 , ... , y KT*Fs , y T*Fs+1 , y T*Fs+2 , ...
  • y K with T*Fs as a period take the original value of the previous periodic segment as the estimated value of the current periodic segment, and measure the residual values from the second to the last periodic segments according to the periodic difference model, and combine the original value of the first periodic segment and the residual values from the second to the last periodic segments to form a new signal segment y 1 , y 2 , ... , y T*Fs , y T*Fs+1 -y 1 , y T*Fs+2 -y 2 , ... , y K -y KT*Fs , the new signal segment is equal to the original signal segment, and then single-point difference is performed on the signal points in the new signal segment to obtain the final signal residual segment.
  • the 8-bit overflow rate of the signal residual fragment measured by combining periodic difference and single-point difference As shown in Figure 5, for a strong power frequency signal (sampling rate of 500Hz), with 10 points as one power frequency cycle, the overflow rate of the original data is 100%, the overflow rate of the single-point differential (i.e., the differential data of the first cycle) is about 47%, and the overflow rate of the period differential + single-point differential is about 22%, which is significantly better than the overflow rate of the single-point differential.
  • the overflow rate of the single-point differential i.e., the differential data of the first cycle
  • the overflow rate of the period differential + single-point differential is about 16%, which is slightly worse than the single-point differential.
  • the combination of period differential and single-point differential has obvious advantages.
  • a 24-bit original signal is used and sliced to obtain a 24-bit signal segment.
  • the residual value of the 24-bit signal segment is measured by combining periodic difference and single-point difference, and the residual value is encoded by quotient and remainder operation.
  • Huffman coding is used for secondary encoding to obtain the final processed signal segment. After quotient and remainder operation encoding, the number of bits occupied by the signal point is still 24 bits.
  • the difference between the second embodiment and the first embodiment is that the residual value of the 24-bit signal segment is measured by a single-point difference method and a combination of single-point differences.
  • the difference between the third embodiment and the first embodiment is that the residual value of the 24-bit signal segment is measured by single-point difference.
  • the comparative example uses a 24-bit original signal, slices the original signal, and obtains a 24-bit signal segment.
  • the residual value of the 24-bit signal segment is measured by single-point differential, and the residual value is encoded by Huffman coding.
  • the secondary coding uses Huffman coding to obtain the final processed signal segment. After the first Huffman coding, the number of bits of the signal point increases from 24 bits to 32 bits, which increases the workload of signal compression.
  • the compression ratio of Example 1 is stable at about 3.47
  • the compression ratio of Example 2 is stable at about 3.29
  • the compression ratio of Example 3 is stable at about 3.24
  • the compression ratio of the comparative example is stable at about 3.02.
  • the signal compression ratios of the three examples are all higher than those of the comparative example, and the compression ratio of Example 1> the compression ratio of Example 2> the compression ratio of Example 3.
  • the storage compression ratio is positively correlated with the length of the signal segment. After the length of the signal segment is greater than 15 seconds, the storage compression ratio tends to be stable.
  • the overflow rate of Example 1 is about 19%
  • the overflow rate of Example 2 is about 40%
  • the overflow rates of Example 3 and the comparative example are about 44%. It can be seen that the 8-bit overflow rate of Example 1 is significantly lower than that of Examples 2, 3 and the comparative example, indicating that the combination of periodic differential + single-point differential can effectively reduce
  • the low signal overflow rate effectively reduces the number of signal encoding times, improves the storage compression ratio and also improves the compression efficiency.
  • the present invention also provides a signal processing model, comprising: an acquisition module for inputting or acquiring signals; a processing module for executing the signal processing method; and a storage module for storing processed signals.
  • the signal detection method, signal processing method and signal processing model of the present invention obtain the processing value of the signal point through signal preprocessing; detect whether the processing value exceeds the set overflow threshold range; select different reprocessing methods according to the detection result of the processing value, which can effectively process the original signal, further improve the signal compression ratio and compression efficiency, and significantly save storage space, thereby saving storage costs for enterprises, and has high application value.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

本发明涉及信号检测技术领域，具体公开了一种信号检测方法、信号处理方法、信号处理模型。信号检测方法包括：信号预处理，即对任一信号片段中的信号点进行处理，以得到信号点的处理值；检测所述处理值是否超出设定的溢出阈值范围，以根据处理值的检测结果选择不同的再处理方式。尤其是在信号再处理之前的检测过程，可以将信号根据处理值的检测结果针对性处理，有效提高了信号的处理效果。

Description

信号检测方法、信号处理方法、信号处理模型 技术领域
本发明涉及信号检测技术领域，尤其涉及一种信号检测方法、信号处理方法、信号处理模型。
背景技术
对于采用单相电源的模拟前端，采集到的模拟信号存在直流偏置，大幅提升了原始数据的动态范围，进而需要更大位宽的ADC完成高保真的模数转换。现有的信号传输和存储方式没有对信号进行检测或分期处理，导致ADC位宽较大，对信号传输与存储的要求较高，势必增加硬件成本。
发明内容
本发明要解决的技术问题是：提供一种信号检测方法、信号处理方法、信号处理模型。
本发明解决其技术问题所采用的技术方案是：所述信号检测方法包括：信号预处理，即对任一信号片段中的信号点进行处理，以得到信号点的处理值；检测所述处理值是否超出设定的溢出阈值范围，以根据处理值的检测结果选择不同的再处理方式。
进一步的，所述处理值为滤波处理值或通过信号预处理模型测量信号点的残差值；所述信号预处理模型包括：AR模型、单点差分模型、周期差分模型、单点差分与周期差分的结合模型中的任一种；其中当信号无周期性干扰且采样率较高时，采用单点差分模型测量信号点的残差值；当信号存在周期性干扰时，采用周期差分模型测量信号点的残差值；当信号存在周期性干扰且采样率较高时，采用单点差分与周期差分的结合模型测量信号点的残差值。
进一步的，根据单点差分模型测量信号点的残差值包括：残差值单点差分＝当前信号点的原始值-上一信号点的原始值；根据周期差分模型测量信号点的残差值包括：残差值周期差分＝当前周期片段的原始值-J*当前周期片段的估计值，其中，所述当前周期片段的估计值＝上一周期片段的原始值，J表示系数；通过单点差分与周期差分的结合模型测量信号点的残差值包括：将所述信号点组合成多个周期片段测量周期差分的残差值，即残差值结合差分＝残差值周期差分(k)-残差值周期差分(k-1)，其中k＝2，3，...，K。
进一步的，所述信号检测方法还包括在信号预处理之前对信号片段进行分期检测；所述分期检测包括：滑窗处理；检测窗内信号的波动性；根据窗内信号的波动性结果检测信号分期。
进一步的，所述窗内信号的波动性结果的表征方式包括窗内信号的方差和均值、窗内 信号的归一化线长中的任一种；其中根据窗内信号的方差和均值检测信号分期包括：获取窗内信号的方差和均值；检测方差不小于第二阈值的信号片段为波动期，检测方差小于第二阈值的信号片段为平稳期；通过窗内信号的归一化线长检测信号分期包括：获取归一化线长，即测量窗内信号的一阶差分的绝对值的平均值，其中A是窗内信号点的个数，x(a)表示窗内第a个信号点，a＝1,2,...,A-1；检测归一化线长超过溢出阈值的信号片段为波动期，检测归一化线长不超过溢出阈值的信号片段为平稳期。
本发明还提供了一种基于如前所述的信号检测方法的信号处理方法，所述根据处理值的检测结果选择不同的再处理方式包括：当任一信号点的处理值超出设定的溢出阈值范围时，对信号点的处理值进行映射编码；其中所述映射编码包括：对信号点的处理值重复执行N次映射运算至在溢出阈值范围内，得到映射标记和映射值，则该信号点编码为：N个映射标记+M种映射值；其中M为映射值的类型，M≥1。
进一步的，当M＝1时，所述映射运算包括减法运算；所述减法运算为将信号点的处理值减去所述溢出阈值；所述映射值为最后一次减法运算得到的差值，则该信号点编码为：N个映射标记+差值。
进一步的，当M＝2时，所述映射运算包括对数运算、商余运算中的任一种；其中所述对数运算为Z＝log2(D)，D为处理值；所述映射值为最后一次对数运算得到的浮点值Z，其整数段为int(Z)，小数段为Z-int(Z)；则该信号点编码为：N个映射标记+int(Z)+Z-int(Z)；所述商余运算为将信号点的处理值除以所述溢出阈值，所述映射值为最后一次商余运算得到的商值Q和N次商余运算得到的N个余数C，则该信号点编码为：N个映射标记+商值Q+N个余数C。
进一步的，所述溢出阈值范围由溢出阈值决定，溢出阈值为-2n-1或2n-1,n为期望编码位宽，所述溢出阈值范围为(-2n-1,2n-1)；设定信号片段的n位溢出率等于在任一信号片段中处理值超过溢出阈值范围的信号点个数除以信号片段中总的信号点个数；当信号片段处于平稳期且信号片段的n位溢出率低于第一阈值时，选择采用减法运算的映射编码对处理值进行编码；当信号片段处于波动期且信号片段的平均信息熵小于设定熵阈值时，选择采用对数运算的映射编码对处理值进行编码；当信号片段处于波动期且信号片段的平均信息熵不小于设定熵阈值时或信号片段处于平稳期且信号片段的n位溢出率高于第一阈值时，选择采用商余运算的映射编码对处理值进行编码。
进一步的，所述根据处理值的检测结果选择不同的再处理方式还包括：对映射编码后 的信号点或处理值不超出设定的溢出阈值范围的信号点进行二次编码；所述二次编码包括：霍夫曼编码及其变体编码、算术编码及其变体编码、区间编码及其变体编码，非对称数字系统编码中的至少一种。
本发明还提供了一种信号处理模型，包括：采集模块，用于采集信号；处理模块，用于执行如前所述的信号处理方法；存储模块，用于存储处理后的信号。
本发明的有益效果是，本发明的信号处理方法及信号处理模型先对信号进行信号预处理，得到信号点的处理值；然后检测所述处理值是否超出设定的溢出阈值范围；再根据处理值的检测结果选择不同的再处理方式，有效的将信号根据处理值的检测结果做了针对性区分处理，有效提高了信号的处理效果。
附图说明
下面结合附图和实施例对本发明进一步说明。
图1是本发明的信号处理方法的流程图。
图2是本发明的信号处理方法的具体工作过程图。
图3是本发明的商余运算编码的一种示意图。
图4是本发明的周期差分和单点差分运算的示意图。
图5是本发明的不同信号预处理模型的溢出率的一对比图。
图6是本发明的不同信号预处理模型的溢出率的另一对比图。
图7是本发明的实施例一的处理过程图。
图8是本发明的实施例二的处理过程图。
图9是本发明的实施例三的处理过程图。
图10是对比例的处理过程图。
图11是本发明的实施例与对比例的压缩比变化的曲线图。
图12是本发明的实施例与对比例的溢出率变化的曲线图。
具体实施方式
现在结合附图对本发明作进一步详细的说明。这些附图均为简化的示意图，仅以示意方式说明本发明的基本结构，因此其仅显示与本发明有关的构成。
在本发明的描述中，需要理解的是，术语“中心”、“纵向”、“横向”、“长度”、“宽度”、“厚度”、“上”、“下”、“前”、“后”、“左”、“右”、“竖直”、“水平”、“顶”、“底”“内”、“外”、“顺时针”、“逆时针”、“轴向”、“径向”、“周向”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本 发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。此外，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征。在本发明的描述中，除非另有说明，“多个”的含义是两个或两个以上。
在本发明的描述中，需要说明的是，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或一体地连接；可以是机械连接，也可以是电连接；可以是直接相连，也可以通过中间媒介间接相连，可以是两个元件内部的连通。对于本领域的普通技术人员而言，可以具体情况理解上述术语在本发明中的具体含义。
如图1至图2所示，本发明的信号处理方法包括以下步骤：先采用信号检测方法检测信号的处理值是否超出设定的溢出阈值范围，再根据处理值的检测结果选择不同的再处理方式。其中信号检测方法包括：S1、信号预处理，即对任一信号片段中的信号点进行处理，以得到信号点的处理值；S2、检测所述处理值是否超出设定的溢出阈值范围；以根据处理值的检测结果选择不同的再处理方式。根据处理值的检测结果选择不同的再处理方式包括：当任一信号点的处理值超出设定的溢出阈值范围时，先对信号点的处理值进行映射编码，再进行二次编码；当任一信号点的处理值不超出设定的溢出阈值范围，对其处理值直接进行二次编码；其中所述映射编码包括：对信号点的处理值重复执行N次映射运算至在溢出阈值范围内，得到映射标记和映射值，则该信号点编码为：N个映射标记+M种映射值；其中M为映射值的类型，M≥1。
例如图2所示，M-bit的原始信号经过滑窗处理可以获得多个M-bit的信号片段，对该信号片段进行分期检测后，利用信号预处理模型计算出信号点的处理值，再判断信号点的处理值是否超出溢出阈值范围，若是，则进行映射编码处理，若否，则保存该处理值，然后将映射编码后的处理值以及直接保存的处理值均进行二次编码后得到最终的信号。
可选的，所述二次编码包括但不限于：霍夫曼编码及其变体编码、算术编码及其变体编码、区间编码及其变体编码，非对称数字系统编码中的至少一种。也就是说，进行信号编码后的信号可以再进行二次编码，更进一步压缩信号。
需要说明的是，本发明所说的信号可以任意类型的信号，例如但不限于脑电信号、肌电信号、心电信号等生理信号，也可以是图像信号、软件数据等可存储的数据。在获取原始信号后可以先对原始信号进行切片处理，划分为多个信号片段，然后将信号片段分期检测，根据波动性标记信号分期，然后对每一信号片段中的信号点进行信号预处理和信号编码，得 到压缩信号。其中，处理值为通过信号预处理模型测量得到的信号点的残差值或滤波处理值。对信号片段中的所有信号点的处理值进行判断，如果某一信号点的处理值未超过设定的溢出阈值范围，则不做编码处理；如果某一信号点的处理值超过了设定的溢出阈值范围，则需要进行信号编码，以减少信号占用空间。信号点的编码方式为“N个映射标记+M种映射值”，其中映射标记是为了便于后续能够对压缩信号进行解码，映射值为至少一种。
例如，当M＝1时，映射运算包括减法运算。减法运算为将信号点的处理值减去溢出阈值，映射值为最后一次减法运算得到的差值，则该信号点编码为：N个映射标记+差值。也就是说，当M＝1时，映射值为减法的差值，信号点的处理值减去溢出阈值，得到第一次减法的差值，若第一次减法的差值仍然超出溢出阈值范围，则继续对差值做减法运算，直至差值在溢出阈值范围内。在进行编码时，映射标记可以采用字符，N表示做减法的次数，映射值选取最后一次做减法得到的差值。
例如，当M＝2时，映射运算包括对数运算、商余运算中的任一种。其中，对数运算为Z＝log2(D)，D为处理值；映射值为最后一次对数运算得到的浮点值Z，其整数段为int(Z)，小数段为Z-int(Z)；则该信号点编码为：N个映射标记+int(Z)+Z-int(Z)。商余运算为将信号点的处理值除以溢出阈值，映射值为最后一次商余运算得到的商值Q和N次商余运算得到的N个余数C，则该信号点编码为：N个映射标记+商值Q+N个余数C。也就是说，在进行对数运算时，对信号点的处理值做对数运算后得到的浮点值如果仍然超出溢出阈值范围，则继续对浮点值做对数运算，直至浮点值在溢出阈值范围内。进行编码时，映射标记可以采用字符，N表示做对数运算的次数，映射值包含整数段int(Z)和小数段Z-int(Z)两种，并且，编码时，整数段int(Z)仅仅占用1个字节，小数段Z-int(Z)可以占用多个字节。在进行商余运算时，对信号点的处理值做除法，即将处理值除以溢出阈值，可以得到商值Q和余数C，如果商值Q仍然超出溢出阈值范围，则继续将商值Q除以溢出阈值，直至商值Q在溢出阈值范围内。进行编码时，映射标记可以采用字符，N表示做商余运算的次数，映射值包含商值Q和余数C两种，商值Q选取最后一次除法得到的商值，余数C则保留全部。商余运算通过做除法将信号点映射为更小的商值和余数。
需要说明的是，上述三种运算中N的数值可能相同也可能不同。本发明的编码方式可以在保留信号特征的同时，尽可能地压缩信号的数据量，进一步提高压缩比。
在本发明中，溢出阈值范围由溢出阈值决定，溢出阈值为-2n-1或2n-1,n为期望编码位宽，则溢出阈值范围为(-2n-1,2n-1)，溢出阈值范围不包含端点。当溢出阈值参与映射运算时，其值取决于计算机有符号表示的数值中绝对值最大的数是正数还是负数，换句话说，其值取 决于计算机n-bit有符号信号所能表示信号范围，若为(-2n-1,2n-1]，则溢出阈值取2n-1；若为[-2n-1,2n-1)，则溢出阈值取-2n-1，但同一计算机进行信号压缩时应统一为正数或者负数中的一种。通过映射编码对处理值进行编码包括：设定信号片段的n位溢出率等于在任一信号片段中处理值超过溢出阈值范围的信号点个数除以信号片段中总的信号点个数；当信号片段处于平稳期且信号片段的n位溢出率低于第一阈值时，选择M＝1的映射编码对处理值进行编码；当信号片段处于波动期且信号片段的平均信息熵小于设定熵阈值时，选择M＝2映射编码的对数运算对处理值进行编码；当信号片段处于波动期，且信号片段的平均信息熵不小于设定熵阈值时或信号片段处于平稳期且信号片段的n位溢出率高于第一阈值时，选择M＝2映射编码的商余运算对处理值进行编码。例如，第一阈值为1-10％，优选为3％、5％、7％。也就是说，在选取信号编码方式时，可以根据信号片段中信号点的n位溢出率、信号波动状态(平稳期或波动期)和信号片段的平均信息熵来确定选择哪种编码方式。根据信号的特点选择合适的编码方式，例如处于平稳期的信号的溢出率小，就没有必要用计算量相对较大的对数运算、商余运算了，既能够对信号进行更好地压缩，节省测量资源，提高效率，还能够保留信号中的重要特征。
例如图3所示，n＝8，溢出阈值为-128或128，溢出阈值范围为(-128，128)，设原始信号点为24bit，信号片段包含10个信号点，对信号片段中的信号点做单点差分，得到9个残差值，溢出率为50％。因此可以选择商余运算对处理后的信号片段内的10个信号点(1个原始值+9个残差值)进行编码。进行商余运算时，溢出阈值可以设为128，第1个信号点为15000，将15000÷128＝117(在溢出阈值范围内)，余数为24，因此，对信号点15000的编码为“S+117+24”，S为映射标记；将信号点16800÷128＝131，余数为32，商值131仍然超出了溢出阈值范围，因此继续对商值131做除法，131÷128＝1(在溢出阈值范围内)，余数为3，因此，对信号点16800的编码为“S+S+1+3+32”。通过商余运算可以得到信号片段内所有信号点的编码(如图3所示)。经过信号编码后，10个信号从原来占用30个字节降低至22个字节，有效减少了信号量，节省了存储空间。并且，商余运算是一种无损压缩，编码后的信号位数与原始信号位数相等，这样解码后仍然可以获取有效的信号(即保留了原始特征)进行后续的分析处理。
例如，信号检测方法还包括在信号预处理之前对信号片段进行分期检测；所述分期检测包括：滑窗处理；检测窗内信号的波动性；根据窗内信号的波动性结果检测信号分期。也就是说，在进行信号预处理之前可以先对信号片段进行分类，确定该信号片段是处于平稳期还是波动期，然后根据信号片段的分类结果再进行信号预处理和选择性的再处理。这样，可 以进一步提升信号的处理效果，即信号处理后的压缩效果以及压缩效率。
所述窗内信号的波动性的表征方式包括窗内信号的方差和均值、窗内信号的归一化线长中的任一种；其中根据窗内信号的方差和均值检测信号分期包括：获取窗内信号的方差和均值；检测方差不小于第二阈值的信号片段为波动期，检测方差小于第二阈值的信号片段为平稳期。所述第二阈值为窗内信号的方差均值的2-10倍，优选为3倍、5倍、8倍。
通过窗内信号的归一化线长检测信号分期包括：获取归一化线长，即测量窗内信号的一阶差分的绝对值的平均值：其中，A是窗内信号点的个数，x(a)表示窗内第a个信号点，a＝1,2,...,A-1；检测归一化线长超过溢出阈值的信号片段为波动期，检测归一化线长不超过溢出阈值的信号片段为平稳期。
具体的，在本发明中，当处理值为滤波处理值时，其滤波值的处理方式主要采用低通滤波或高通滤波，来消除与目标信号无关的背景噪声。当处理值为残差值时，其主要采用的信号预处理模型包括：AR模型、单点差分模型、周期差分模型、单点差分与周期差分的结合模型中的任一种。其中，当信号无周期性干扰且采样率较高时，采用单点差分模型测量信号点的残差值；当信号存在周期性干扰时，采用周期差分模型测量信号点的残差值；当信号存在周期性干扰且采样率较高时，采用单点差分与周期差分的结合模型测量信号点的残差值。
例如，AR模型的表达式为：
其中，t＝P+1，...，K，为P阶中心化AR模型的系数，为原始信号与其估计值之间的残差。对于信号压缩而言，获取以最小化残差ε[t]是提升压缩比的关键。信号片段的残差矢量为：
最小化AR模型的系数可以表示为：
例如，另P＝1，上述P阶中心化AR模型可以表示为：
ε[t]＝y[t]-y[t-1]，t＝2，...，K，
即获得单点差分模型。单点差分模型能降低残差值的溢出率。
例如，令K＝2P且P＝T*FS，则上述P阶AR模型可以表示为：
ε[t]＝y[t]-ω·y[t-P]，t＝P+1，...，2P，
即获得周期差分模型，其中，T表示干扰的最大周期，Fs为采样率，w为趋势补偿系数，能够反映前后周期整体复制的变化率。周期差分模型可以去除原始信号中的周期干扰，从而降低残差值的溢出率。
单点差分模型和周期差分模型无需动态求解系数测量效率可以明显提高，由于省去了系数的求解，还能够节约存储空间。在实际应用时，可以根据信号是否存在周期性干扰和采样率来选取信号预处理模型。
具体的，根据单点差分模型测量信号点的残差值包括：残差值单点差分＝当前信号点的原始值-上一信号点的原始值。根据周期差分模型测量信号点的残差值包括：残差值周期差分＝当前周期片段的原始值-J*当前周期片段的估计值，其中，当前周期片段的估计值＝上一周期片段的原始值，J表示系数。通过单点差分与周期差分的结合模型测量信号点的残差值包括：将信号点组合成多个周期片段测量周期差分的残差值，即残差值结合差分＝残差值周期差分(k)-残差值周期 差分(k-1)，其中k＝2，3，...，K。
例如图4所示，T＝1/50或1/60(工频干扰的最大周期)，干扰最大周期对应的信号点数为T*Fs。设某一信号片段包含的信号点为y1，y2，...，yK，共K个信号点，将K个信号点以T*Fs为一个周期划分为多个周期片段y1，y2，...，yK-T*Fs、yT*Fs+1，yT*Fs+2，...，yK，以上一周期片段的原始值作为当前周期片段的估计值，根据周期差分模型测量第二个至最后一个周期片段的残差值，将第一个周期片段的原始值和第二个至最后一个周期片段的残差值组成新的信号片段y1，y2，...，yT*Fs，yT*Fs+1-y1，yT*Fs+2-y2，...，yK-yK-T*Fs，新的信号片段与原始的信号片段等长，然后对新的信号片段内的信号点做单点差分，得到最终的信号残差片段。
相比原始信号，通过周期差分和单点差分结合测量得到的信号残差片段的8-bit溢出率 显出降低。如图5所示，对于强工频信号(采样率为500Hz)，以10个点数为一个工频周期，原始数据的溢出率为100％，单点差分(即第一个周期的差分数据)的溢出率约为47％，而周期差分+单点差分的溢出率约为22％，明显优于单点差分的溢出率。如图6所示，对于弱工频信号(采样率为1000Hz)，以20个点数为一个工频周期，单点差分(即第一个周期的差分数据)的溢出率约为12％，而周期差分+单点差分的溢出率约为16％，稍劣于单点差分。对于50Hz/60Hz的工频干扰来说，周期差分与单点差分结合的方式具有明显优势。
下面通过具体的实施例来说明本发明的有益效果。
实施例一
如图7所示，采用24bit原始信号，对原始信号进行切片处理，得到24bit信号片段。通过周期差分和单点差分结合的方式测量24bit信号片段的残差值，对残差值采用商余运算进行编码，二次编码选用霍夫曼编码，得到最终的处理后信号片段。商余运算编码后，信号点的占用位数仍然是24位。
实施例二
如图8所示，实施例二与实施例一的区别在于，通过单点差分和单点差分结合的方式测量24bit信号片段的残差值。
实施例三
如图9所示，实施例三与实施例一的区别在于，通过单点差分测量24bit信号片段的残差值。
对比例
如图10所示，对比例(即现有技术)采用24bit原始信号，对原始信号进行切片处理，得到24bit信号片段。通过单点差分测量24bit信号片段的残差值，对残差值采用霍夫曼编码进行编码，二次编码选用霍夫曼编码，得到最终的处理后信号片段。第一次霍夫曼编码后，信号点的位数从24bit增加到了32bit，给信号压缩反而增加了工作量。
从图11可知，实施例一的压缩比稳定在3.47左右，实施例二的压缩比稳定在3.29左右，实施例三的压缩比稳定在3.24左右，而对比例的压缩比稳定在3.02左右。将实施例一至实施例三以及对比例进行比较，三个实施例的信号压缩比均高于对比例，并且，实施例一的压缩比>实施例二的压缩比>实施例三的压缩比。表明本发明的信号处理方法相比现有技术能够将信号的存储压缩比提升7.2％～15％。并且，存储压缩比与信号片段的长度呈正相关，信号片段长度大于15秒后，存储压缩比趋向平稳。如图12所示，实施例一的溢出率约为19％，实施例二的溢出率约40％，实施例三和对比例的溢出率约为44％。由此可知，实施例一的8bit溢出率明显低于实施例二、三和对比例，表明采用周期差分+单点差分结合的方式能够有效降 低信号的溢出率，进而有效减少信号编码的次数，提高存储压缩比的同时还能够提升压缩效率。
本发明还提供了一种信号处理模型，包括：采集模块，用于输入或采集信号；处理模块，用于执行所述信号处理方法；存储模块，用于存储处理后的信号。
综上所述，本发明的信号检测方法、信号处理方法及信号处理模型，通过信号预处理得到信号点的处理值；检测所述处理值是否超出设定的溢出阈值范围；根据处理值的检测结果选择不同的再处理方式，能够对原始信号进行有效处理，进一步提升信号压缩比和压缩效率，显著节约存储空间，从而也能够为企业节约存储成本，具有很高的应用价值。
以上述依据本发明的理想实施例为启示，通过上述的说明内容，相关工作人员完全可以在不偏离本项发明技术思想的范围内，进行多样的变更以及修改。本项发明的技术性范围并不局限于说明书上的内容，必须要如权利要求范围来确定其技术性范围。

Claims (11)

  1. 一种信号检测方法，其特征在于，包括：
    信号预处理，即对任一信号片段中的信号点进行处理，以得到信号点的处理值；
    检测所述处理值是否超出设定的溢出阈值范围，以根据处理值的检测结果选择不同的再处理方式。
  2. 如权利要求1所述的信号检测方法，其特征在于，
    所述处理值为滤波处理值或通过信号预处理模型测量信号点的残差值；
    所述信号预处理模型包括：AR模型、单点差分模型、周期差分模型、单点差分与周期差分的结合模型中的任一种；其中
    当信号无周期性干扰且采样率较高时，采用单点差分模型测量信号点的残差值；
    当信号存在周期性干扰时，采用周期差分模型测量信号点的残差值；
    当信号存在周期性干扰且采样率较高时，采用单点差分与周期差分的结合模型测量信号点的残差值。
  3. 如权利要求2所述的信号检测方法，其特征在于，
    根据单点差分模型测量信号点的残差值包括：残差值单点差分＝当前信号点的原始值-上一信号点的原始值；
    根据周期差分模型测量信号点的残差值包括：残差值周期差分＝当前周期片段的原始值-J*当前周期片段的估计值，其中，所述当前周期片段的估计值＝上一周期片段的原始值，J表示系数；通过单点差分与周期差分的结合模型测量信号点的残差值包括：将所述信号点组合成多个周期片段测量周期差分的残差值，即残差值结合差分＝残差值周期差分(k)-残差值周期差分(k-1)，其中k＝2，3，...，K。
  4. 如权利要求1所述的信号检测方法，其特征在于，
    所述信号检测方法还包括在信号预处理之前对信号片段进行分期检测；
    所述分期检测包括：
    滑窗处理；
    检测窗内信号的波动性；
    根据窗内信号的波动性结果检测信号分期。
  5. 如权利要求4所述的信号检测方法，其特征在于，
    所述窗内信号的波动性结果的表征方式包括窗内信号的方差和均值、窗内信号的归一化线长中的任一种；其中
    根据窗内信号的方差和均值检测信号分期包括：
    测量窗内信号的方差和均值；检测方差不小于第二阈值的信号片段为波动期，检测方差小于第二阈值的信号片段为平稳期；
    通过窗内信号的归一化线长检测信号分期包括：
    获取归一化线长，即测量窗内信号的一阶差分的绝对值的平均值，
    其中A是窗内信号点的个数，x(a)表示窗内第a个信号点，a＝1,2,...,A-1；
    检测归一化线长超过溢出阈值的信号片段为波动期，检测归一化线长不超过溢出阈值的信号片段为平稳期。
  6. 一种基于如权利要求1-5任一项所述的信号检测方法的信号处理方法，其特征在于，
    所述根据处理值的检测结果选择不同的再处理方式包括：当任一信号点的处理值超出设定的溢出阈值范围时，对信号点的处理值进行映射编码；其中
    所述映射编码包括：
    对信号点的处理值重复执行N次映射运算至在溢出阈值范围内，得到映射标记和映射值，则该信号点编码为：N个映射标记+M种映射值；其中M为映射值的类型，M≥1。
  7. 如权利要求6所述的信号处理方法，其特征在于，
    当M＝1时，所述映射运算包括减法运算；
    所述减法运算为将信号点的处理值减去所述溢出阈值；所述映射值为最后一次减法运算得到的差值，则该信号点编码为：N个映射标记+差值。
  8. 如权利要求6所述的信号处理方法，其特征在于，
    当M＝2时，所述映射运算包括对数运算、商余运算中的任一种；其中
    所述对数运算为Z＝log2(D)，D为处理值；所述映射值为最后一次对数运算得到的浮点值Z，其整数段为int(Z)，小数段为Z-int(Z)；则该信号点编码为：N个映射标记+int(Z)+Z-int(Z)；
    所述商余运算为将信号点的处理值除以所述溢出阈值，所述映射值为最后一次商余运算得到的商值Q和N次商余运算得到的N个余数C，则该信号点编码为：N个映射标记+商值Q+N个余数C。
  9. 如权利要求7或8所述的信号处理方法，其特征在于，
    所述溢出阈值范围由溢出阈值决定，溢出阈值为-2n-1或2n-1,n为期望编码位宽，所述溢出阈值范围为(-2n-1,2n-1)；
    设定信号片段的n位溢出率等于在任一信号片段中处理值超过溢出阈值范围的信号点个数除 以信号片段中总的信号点个数；
    当信号片段处于平稳期且信号片段的n位溢出率低于第一阈值时，选择采用减法运算的映射编码对处理值进行编码；
    当信号片段处于波动期且信号片段的平均信息熵小于设定熵阈值时，选择采用对数运算的映射编码对处理值进行编码；
    当信号片段处于波动期且信号片段的平均信息熵不小于设定熵阈值时或信号片段处于平稳期且信号片段的n位溢出率高于第一阈值时，选择采用商余运算的映射编码对处理值进行编码。
  10. 如权利要求6所述的信号处理方法，其特征在于，
    所述根据处理值的检测结果选择不同的再处理方式还包括：对映射编码后的信号点或处理值不超出设定的溢出阈值范围的信号点进行二次编码；
    所述二次编码包括：霍夫曼编码及其变体编码、算术编码及其变体编码、区间编码及其变体编码，非对称数字系统编码中的至少一种。
  11. 一种信号处理模型，其特征在于，包括：
    处理模块，用于执行如权利要求6-10任一项所述的信号处理方法；
    存储模块，用于存储处理后的信号。
PCT/CN2023/093732 2022-10-31 2023-05-12 信号检测方法、信号处理方法、信号处理模型 WO2024093188A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211342078.2A CN115409074B (zh) 2022-10-31 2022-10-31 信号检测方法、信号处理方法、信号处理模型
CN202211342078.2 2022-10-31

Publications (1)

Publication Number Publication Date
WO2024093188A1 true WO2024093188A1 (zh) 2024-05-10

Family

ID=84168118

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/093732 WO2024093188A1 (zh) 2022-10-31 2023-05-12 信号检测方法、信号处理方法、信号处理模型

Country Status (2)

Country Link
CN (1) CN115409074B (zh)
WO (1) WO2024093188A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115409074B (zh) * 2022-10-31 2023-03-24 博睿康科技（常州）股份有限公司 信号检测方法、信号处理方法、信号处理模型

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090313188A1 (en) * 2008-06-13 2009-12-17 James Bryan Baker Computationally Efficient Signal Classifier
CN109700450A (zh) * 2018-12-28 2019-05-03 联想(北京)有限公司 一种心率检测方法及电子设备
CN111513706A (zh) * 2020-04-20 2020-08-11 重庆邮电大学 一种针对含有异常r波的心电信号的检测方法和装置
CN112347413A (zh) * 2020-11-06 2021-02-09 哲库科技(北京)有限公司 信号处理方法、信号处理器、设备及存储介质
CN115409074A (zh) * 2022-10-31 2022-11-29 博睿康科技（常州）股份有限公司 信号检测方法、信号处理方法、信号处理模型

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI494081B (zh) * 2012-03-16 2015-08-01 Univ Nat Cheng Kung 心電圖訊號壓縮系統及解壓縮系統
CN106714036B (zh) * 2016-12-15 2019-06-18 成都航天通信设备有限责任公司 一种数字通讯控制方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090313188A1 (en) * 2008-06-13 2009-12-17 James Bryan Baker Computationally Efficient Signal Classifier
CN109700450A (zh) * 2018-12-28 2019-05-03 联想(北京)有限公司 一种心率检测方法及电子设备
CN111513706A (zh) * 2020-04-20 2020-08-11 重庆邮电大学 一种针对含有异常r波的心电信号的检测方法和装置
CN112347413A (zh) * 2020-11-06 2021-02-09 哲库科技(北京)有限公司 信号处理方法、信号处理器、设备及存储介质
CN115409074A (zh) * 2022-10-31 2022-11-29 博睿康科技（常州）股份有限公司 信号检测方法、信号处理方法、信号处理模型

Also Published As

Publication number Publication date
CN115409074A (zh) 2022-11-29
CN115409074B (zh) 2023-03-24

Similar Documents

Publication Publication Date Title
TWI494081B (zh) 心電圖訊號壓縮系統及解壓縮系統
Blalock et al. Sprintz: Time series compression for the internet of things
US7650249B2 (en) Data compression for a waveform data analyzer
WO2024093188A1 (zh) 信号检测方法、信号处理方法、信号处理模型
TW201322118A (zh) 浮點資料之壓縮
Huang et al. A divide-and-compress lossless compression scheme for bearing vibration signals in wireless sensor networks
Daubechies et al. Beta expansions: a new approach to digitally corrected A/D conversion
JP5220071B2 (ja) 欲張り探索を使用して符号測定値からスパース信号を再構築する方法
CN113873094B (zh) 一种混沌压缩感知图像加密方法
Tsai et al. VLSI implementation of lossless ECG compression algorithm for low power devices
CN116170027B (zh) 一种毒检设备数据管理系统及处理方法
CN100490494C (zh) 解码设备、逆量化方法、分布确定方法及其程序
JP5079110B2 (ja) 圧縮された整数データを記憶・転送するためのシステム
CN104320143B (zh) 三相电压、电流信号波形采样数据的压缩方法及装置
CN106841865B (zh) 短时电能质量扰动信号的单比特采样与重构方法
Huang et al. 2-D compression of ECG signals using ROI mask and conditional entropy coding
CN111220876B (zh) 一种暂态录波数据压缩方法、装置、设备及存储介质
CN110595529B (zh) 一种强背景噪声下微弱周期信号的快速检测方法
CN110727681B (zh) 数据存储方法及装置
JPH11514165A (ja) 損失を伴わない波形データ圧縮装置及び方法
Giorgi Lightweight Lossless Compression for $ N $-Dimensional Data in Multi-Sensor Systems
CN109635781B (zh) 基于小波变换的数字信号粗大数据检测修正方法及系统
EP3824556B1 (en) Systems and methods for low-power encoding of continuous physiological signals in a remote physiological monitor
CN109286817A (zh) 一种视频编码中dct系数的量化失真信息处理方法
Sahasranand et al. Anomaly-aware adaptive sampling for electrical signal compression

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23884130

Country of ref document: EP

Kind code of ref document: A1