WO2016023349A1 - 基于表面肌电信号的动作识别方法和设备 - Google Patents

基于表面肌电信号的动作识别方法和设备 Download PDF

Info

Publication number
WO2016023349A1
WO2016023349A1 PCT/CN2015/073369 CN2015073369W WO2016023349A1 WO 2016023349 A1 WO2016023349 A1 WO 2016023349A1 CN 2015073369 W CN2015073369 W CN 2015073369W WO 2016023349 A1 WO2016023349 A1 WO 2016023349A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
surface electromyogram
sliding
effective surface
electromyogram signal
Prior art date
Application number
PCT/CN2015/073369
Other languages
English (en)
French (fr)
Inventor
丁强
高小榕
林科
Original Assignee
华为技术有限公司
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 华为技术有限公司 filed Critical 华为技术有限公司
Priority to EP15832528.2A priority Critical patent/EP3167799A4/en
Publication of WO2016023349A1 publication Critical patent/WO2016023349A1/zh
Priority to US15/428,230 priority patent/US20170143226A1/en

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
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/389—Electromyography [EMG]
    • A61B5/397—Analysis of electromyograms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/389—Electromyography [EMG]
    • 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
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • 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
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • 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
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7278—Artificial waveform generation or derivation, e.g. synthesising signals from measured signals

Definitions

  • Embodiments of the present invention relate to motion recognition technologies, and in particular, to a motion recognition method and apparatus based on surface electromyography signals.
  • EMG Surface electromyography
  • the amplitude feature of the surface electromyogram signal is used as the identification parameter for the recognition of the limb motion.
  • the method obtains the window sequence of each sliding window by sliding the collected surface electromyogram signal, and then, The magnitude of the window sequence of each window is calculated to obtain an amplitude feature, and the amplitude feature is compared with the amplitude of the surface electromyogram signal corresponding to each limb motion obtained in advance to determine the limb motion corresponding to the amplitude feature.
  • the prior art method has the following problems: the surface electromyography signal is generated due to other actions of the user, and when the user performs other exercises (running, typing, etc.), the generated surface electromyogram signal and the target surface myoelectric signal may occur. Superimposition, resulting in low recognition accuracy.
  • the magnitude of the surface EMG signal is proportional to the amplitude of the user's limb movement. In order to achieve a higher signal-to-noise ratio, the user needs to perform a larger force on the limb movement for a long time. In the case of operation, the user may feel awkward.
  • Embodiments of the present invention provide a motion recognition method and device based on a surface electromyogram signal, which can improve the accuracy of motion recognition based on surface EMG signals.
  • a first aspect of the present invention provides a motion recognition method based on a surface electromyogram signal, including:
  • a limb motion corresponding to a surface electromyogram signal of the plurality of channels is determined according to a frequency of the effective surface electromyogram signal.
  • the determining an effective surface electromyogram signal based on the surface electromyogram signals of the plurality of channels comprises:
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective surface
  • the start time of the EMG signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time and the cutoff of the effective surface EMG signal are intercepted.
  • a surface electromyogram signal of the plurality of channels between times acts as the effective surface myoelectric signal.
  • the preset amplitude is a superposition of surface electromyogram signals of the plurality of channels The average of the absolute values of the amplitude of the surface EMG signal.
  • the determining the effective surface electromyogram The frequency of the signal, including:
  • the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is used as the frequency of the effective surface electromyogram signal.
  • a second aspect of the present invention provides a motion recognition method based on a surface electromyogram signal, including:
  • a limb motion corresponding to a surface electromyogram signal of the plurality of channels is determined based on an amplitude characteristic of the effective surface electromyogram signal and a frequency of the effective surface electromyogram signal.
  • the determining the effective surface electromyogram signal according to the surface electromyogram signals of the plurality of channels comprises:
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective
  • the start time of the surface electromyogram signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time of the effective surface EMG signal is intercepted.
  • a surface electromyogram signal of the plurality of channels between the cutoff times serves as the effective surface electromyogram signal.
  • the preset amplitude is a superposition of surface electromyogram signals of the plurality of channels The average of the absolute values of the amplitude of the surface EMG signal.
  • the determining the effective surface electromyogram The frequency of the signal, including:
  • the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is used as the frequency of the effective surface electromyogram signal.
  • the extracting the amplitude characteristic of the effective surface electromyogram signal comprises:
  • the effective surface electromyogram Amplitude characteristics of the signal and a frequency of the effective surface EMG signal determine limb motion corresponding to surface electromyographic signals of the plurality of channels, including:
  • a limb motion that matches the amplitude characteristic of the effective surface EMG signal acts as a limb motion corresponding to the surface electromyogram signal of the plurality of channels.
  • a third aspect of the present invention provides a motion recognition device based on a surface electromyogram signal, including:
  • An acquisition module for acquiring surface electromyogram signals of the plurality of channels
  • a first determining module configured to determine an effective surface muscle according to surface electromyogram signals of the plurality of channels electric signal
  • a second determining module configured to determine a frequency of the effective surface electromyogram signal
  • an identification module configured to determine a limb motion corresponding to a surface electromyogram signal of the plurality of channels according to a frequency of the effective surface electromyogram signal.
  • the first determining module is specifically configured to:
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective surface
  • the start time of the EMG signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time and the cutoff of the effective surface EMG signal are intercepted.
  • a surface electromyogram signal of the plurality of channels between times acts as the effective surface myoelectric signal.
  • the preset amplitude is a superposition of surface electromyogram signals of the plurality of channels The average of the absolute values of the amplitude of the surface EMG signal.
  • the second determining module is specifically configured to:
  • the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is used as the frequency of the effective surface electromyogram signal.
  • a fourth aspect of the present invention provides a motion recognition device based on a surface electromyogram signal, including:
  • An acquisition module for acquiring surface electromyogram signals of the plurality of channels
  • a first determining module configured to determine an effective surface electromyogram signal according to surface electromyogram signals of the plurality of channels
  • a second determining module configured to determine a frequency of the effective surface electromyogram signal
  • An extraction module configured to extract amplitude characteristics of the effective surface EMG signal
  • an identification module configured to determine a limb motion corresponding to the surface electromyogram signal of the plurality of channels according to the amplitude characteristic of the effective surface electromyogram signal and the frequency of the effective surface electromyogram signal.
  • the first determining module is specifically configured to:
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective
  • the start time of the surface electromyogram signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time of the effective surface EMG signal is intercepted.
  • a surface electromyogram signal of the plurality of channels between the cutoff times serves as the effective surface electromyogram signal.
  • the preset amplitude is a superposition of surface electromyogram signals of the plurality of channels The average of the absolute values of the amplitude of the surface EMG signal.
  • the second determining module is specifically configured to:
  • the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is used as the frequency of the effective surface electromyogram signal.
  • the identifying module is specifically configured to:
  • a limb motion that matches the amplitude characteristic of the effective surface EMG signal acts as a limb motion corresponding to the surface electromyogram signal of the plurality of channels.
  • a fifth aspect of the present invention provides a motion recognition device based on a surface electromyogram signal, including:
  • processors a processor, a memory, and a system bus, wherein the processor and the memory are connected by the system bus and complete communication with each other;
  • the memory is configured to store a computer execution instruction
  • the processor configured to run the computer to execute an instruction, to cause the action recognition device to execute A first aspect of the invention and a method provided by the first to third possible implementations of the first aspect of the invention.
  • a sixth aspect of the present invention provides a motion recognition device based on a surface electromyogram signal, including:
  • processors a processor, a memory, and a system bus, wherein the processor and the memory are connected by the system bus and complete communication with each other;
  • the memory is configured to store a computer execution instruction
  • the processor configured to execute the computer to execute instructions, to cause the motion recognition device to perform the method provided by the second aspect of the present invention and the first to fifth possible implementation manners of the second aspect of the present invention.
  • the method and device for detecting motion based on surface electromyography signals by acquiring surface electromyogram signals of a plurality of channels, determining effective surface electromyogram signals according to surface electromyogram signals of the plurality of channels; and then determining effective surface
  • the frequency of the EMG signal determines the limb motion corresponding to the surface EMG signals of the multiple channels based on the frequency of the effective surface EMG signal. Since the frequency of the surface electromyogram signal is independent of characteristics such as signal intensity, the method of the present embodiment can significantly improve the accuracy of motion recognition based on the surface myoelectric signal. Moreover, with the frequency as the identification feature, the user does not need to perform a large-scale action to bring a better experience to the user.
  • the method of the embodiment of the present invention can also combine the frequency and amplitude characteristics of the surface electromyogram signal to identify the limb movement corresponding to the surface electromyogram signal, which can not only improve the surface electromyography signal recognition accuracy rate, but also increase the limb movement.
  • the type of identification can also combine the frequency and amplitude characteristics of the surface electromyogram signal to identify the limb movement corresponding to the surface electromyogram signal, which can not only improve the surface electromyography signal recognition accuracy rate, but also increase the limb movement. The type of identification.
  • FIG. 1 is a flowchart of a motion recognition method based on a surface electromyogram signal according to Embodiment 1 of the present invention
  • Embodiment 2 is a method for determining a frequency of an effective surface electromyogram signal according to Embodiment 2 of the present invention
  • FIG. 3 is a flowchart of a method for recognizing a motion based on a surface electromyogram according to Embodiment 3 of the present invention
  • Embodiment 4 is a whole of a motion recognition method based on a surface electromyogram signal according to Embodiment 4 of the present invention; block diagram;
  • FIG. 5 is a schematic structural diagram of a motion recognition device based on a surface electromyogram according to Embodiment 5 of the present invention.
  • FIG. 6 is a schematic structural diagram of a motion recognition device based on a surface electromyogram according to Embodiment 6 of the present invention.
  • FIG. 7 is a schematic structural diagram of a motion recognition device based on a surface electromyogram according to Embodiment 7 of the present invention.
  • FIG. 8 is a schematic structural diagram of a motion recognition device based on a surface electromyogram according to Embodiment 8 of the present invention.
  • FIG. 1 is a flowchart of a method for recognizing a motion based on a surface electromyogram according to a first embodiment of the present invention.
  • the method of the embodiment may be performed by a motion recognition device based on a surface electromyogram signal, and the motion recognition based on a surface electromyogram signal
  • the device can be specifically a terminal device such as a smart phone or a tablet computer.
  • the method of this embodiment may include the following steps:
  • Step 101 Acquire surface electromyogram signals of the plurality of channels.
  • the surface electromyography signal of the plurality of channels is acquired by the motion recognition device based on the surface electromyogram signal, specifically: receiving the surface electromyogram signal of the plurality of channels sent by the collecting device, and placing the collecting device on the surface of the muscle group to be collected for surface muscle.
  • the acquisition of the electrical signal because the method of the present embodiment determines the corresponding limb motion according to the frequency of the surface electromyogram signal, therefore, the user is required to repeat the same limb motion at a specific frequency when collecting the surface myoelectric signal.
  • the limb movement can be not only a rhythmic gesture of the forearm of the upper limb of the human body, but also a rhythmic movement of the leg of the lower limb, or even a rhythmic movement of the body such as the neck and the abdomen, where the rhythmic motion is repeated according to a certain frequency.
  • the acquisition device includes a plurality of sensing nodes, and the data collected by each sensing node serves as a channel.
  • the surface EMG signal the acquisition device can be embedded in a wearable device such as a smart watch or a wristband.
  • a wearable device such as a smart watch or a wristband.
  • Step 102 Determine an effective surface electromyogram signal according to surface electromyogram signals of the plurality of channels.
  • the surface EMG signal has a small amplitude and a low signal-to-noise ratio, it is necessary to pre-process the surface EMG signals of the plurality of channels before determining the effective surface EMG signal, and pre-treating the surface EMG signal. : Signal amplification, power frequency filtering, high-pass filtering, etc. for surface EMG signals.
  • the signal at the beginning of the surface EMG signal collected by the acquisition device may have some interference signals.
  • the user does not perform the corresponding limb movement in time after the operation of the collection device.
  • the collection device can still Some weak surface EMG signals are collected. These weak surface EMG signals are interfering signals. Therefore, it is necessary to eliminate possible interference signals to obtain effective surface EMG signals.
  • the effective surface EMG signal can be determined according to the following method:
  • the surface electromyogram signals of the plurality of channels after pretreatment are superimposed, and the superimposed surface electromyogram signal is divided by the number of channels to obtain a single channel surface electromyogram signal.
  • the single-channel surface electromyogram signal is slid at each sliding moment to obtain a window corresponding to each sliding moment, and the window sequence corresponding to each sliding moment is determined.
  • the average value of the absolute values of the amplitudes, wherein the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N consecutive windows corresponding to N-1 sliding moments before the sliding moment, and a total of N consecutive windows, N Is a positive integer greater than or equal to 2.
  • the start time of the single-channel surface EMG signal refers to the time when the channel signal is started, and the single-channel surface EMG signal is slid at each sliding moment, that is, the single-channel surface signal is at the time of each sliding moment.
  • the time difference between two adjacent sliding moments is a sliding interval.
  • the window width is time t
  • the sliding interval is t/2
  • t 1ms (milliseconds)
  • each window sequence includes four windows, there are 50 windows, and the window number is 1, 2, 3, ... 50, then Starting from the start time of the single-channel surface EMG signal, the window corresponding to the first sliding moment is the window No.
  • the window sequence corresponding to the first sliding moment is 0-4ms, a total of 4 windows, and then, The second sliding moment slides the single-channel surface EMG signal backward for 0.5ms to obtain a second window sequence, the second window sequence is 0.5-4.5ms, and so on, to obtain a window sequence corresponding to each sliding moment.
  • the absolute value of the surface electromyogram signal of the single channel in the window sequence is taken, and then, within the window sequence
  • the absolute values of the single-channel surface EMG signals are summed to obtain the average amplitude of the surface EMG signals of the window sequence.
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective surface electromyogram
  • the start time of the signal, the start time of the effective surface EMG signal plus the preset time to obtain the cut-off time of the effective surface EMG signal, and the multiple channels between the start time and the cut-off time of the effective surface EMG signal are intercepted.
  • the treated surface EMG signal acts as an effective surface EMG signal.
  • the average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment determines whether the average amplitude of the surface electromyogram signal of the window sequence corresponding to the sliding moment is less than a preset amplitude, if the sliding time T corresponds.
  • the average amplitude of the surface electromyogram signal of the window sequence is not less than a preset amplitude, that is, the average amplitude of the surface electromyogram signal of the window sequence corresponding to the sliding time T is greater than or equal to the preset amplitude, and the window sequence corresponding to the sliding time T is
  • the start time is the starting time of the effective surface EMG signal, and the start time of the effective surface EMG signal plus the preset time to obtain the effective surface EMG signal cut-off time.
  • the surface electromyogram signals of the plurality of channels between the start time and the cut-off time are taken as effective surface electromyogram signals.
  • the preset amplitude may be an average value of the absolute values of the amplitudes of the surface electromyogram signals after the superposition of the surface electromyogram signals of the plurality of channels.
  • Step 103 Determine the frequency of the effective surface myoelectric signal.
  • the first method is to calculate the correlation coefficient between the effective surface EMG signal and multiple sine and cosine matrices, assuming the frequency of possible limb movements. For f1,...,fn, then select the sine and cosine matrix with f1,...,fn as the fundamental frequency, and perform the Canonical Correlation Analysis (CCA) calculation on the effective surface EMG signal and each sine and cosine matrix respectively.
  • CCA Canonical Correlation Analysis
  • the maximum correlation coefficient between the effective surface EMG signal and each sine and cosine matrix is taken as the frequency of the effective surface EMG signal.
  • the second method performs Fast Fourier Ttransform (FFT) on the effective surface signal, and determines the frequency distribution of the effective surface EMG signal according to the transformation result, thereby obtaining effective surface muscle communication.
  • FFT Fast Fourier Ttransform
  • the frequency of the number is calculated according to the time interval between two zero crossings of the effective surface electromyogram signal.
  • Step 104 Determine a limb motion corresponding to a surface electromyogram signal of the plurality of channels according to a frequency of the effective surface electromyogram signal.
  • the limb movement corresponding to the 1HZ frequency is “hand fist”
  • the limb movement corresponding to the 2HZ frequency is “elbow bending”
  • the limb movement corresponding to the 3HZ frequency is “ok gesture”
  • the surface electromyogram signal of the plurality of channels is obtained, and the effective surface electromyogram signal is determined according to the surface electromyogram signals of the plurality of channels; then, the frequency of the effective surface electromyogram signal is determined, and finally, according to the surface electromyogram signal
  • the frequency determines the limb motion corresponding to the surface EMG signals of multiple channels. Since the frequency of the surface electromyogram signal is independent of characteristics such as signal intensity, the method of the present embodiment can significantly improve the accuracy of motion recognition based on the surface myoelectric signal. Moreover, with the frequency as the identification feature, the user does not need to perform a large-scale action to bring a better experience to the user.
  • the method of the embodiment has good anti-interference performance, can effectively resist the electromyographic noise interference caused by irrelevant actions, and the user can use under non-stationary conditions such as running, driving, and housework. It is not affected by factors such as changes in skin moisture, changes in electrode contact, and muscle fatigue during use, and has high stability. Users are easy to use, no need to collect user training data in advance, and no need to retrain before each use.
  • FIG. 2 is a method for determining the frequency of the effective surface electromyogram according to the second embodiment of the present invention. As shown in FIG. 2, the embodiment of the present invention is as shown in FIG. The method includes the following steps:
  • Step 201 Calculate correlation coefficients of the effective surface electromyogram signal and the plurality of sine and cosine matrices respectively, wherein the sine and cosine matrix is composed of a sine function of a fundamental frequency and a frequency doubling and a cosine function, and a fundamental frequency of each sine and cosine matrix is different.
  • the sine and cosine matrix with f1,...,fn as the fundamental frequency is chosen as the control.
  • the effective surface EMG signals are respectively subjected to CCA calculations with these sine and cosine matrices to obtain the correlation coefficients between the effective surface EMG signals and the respective sine and cosine matrices.
  • f1 is the fundamental frequency.
  • C xy represents the cross-correlation matrix of x
  • C xx represents the autocorrelation matrix of x
  • Cyy represents the autocorrelation matrix of y.
  • the first step is to construct the following Lagrangian operator by the Lagrangian algorithm:
  • w x can be obtained, and according to the formula 1, w y can be obtained. It can be seen from the formula 2 that the process of finding w x according to the formula 2 is converted into the eigenvalue decomposition problem, and w x and w y are substituted into the ⁇ . The correlation coefficient is obtained in the definition.
  • Step 202 Determine whether a maximum correlation coefficient of the correlation coefficient between the effective surface electromyogram signal and the plurality of sine and cosine matrices is greater than a preset correlation coefficient.
  • the maximum correlation coefficient is found to determine whether the maximum correlation coefficient is greater than the preset correlation coefficient.
  • Step 203 If the maximum correlation coefficient is greater than the preset correlation coefficient, the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is taken as the frequency of the effective surface electromyogram signal.
  • the effective surface EMG signal has the frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient, and the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is taken as the frequency of the effective surface EMG signal.
  • FIG. 3 is a flowchart of a method for recognizing a motion based on a surface electromyogram according to a third embodiment of the present invention.
  • the method of this embodiment differs from the first embodiment in that: in this embodiment, the frequency of the surface electromyogram signal is combined.
  • the amplitude feature determines the limb motion corresponding to the surface myoelectric signal.
  • Step 301 Acquire a surface electromyogram signal of the plurality of channels.
  • Step 302 Determine an effective surface electromyogram signal according to surface electromyogram signals of the plurality of channels.
  • the effective surface EMG signal is determined according to the surface electromyogram signals of the plurality of channels, specifically: superimposing the surface electromyogram signals of the plurality of channels, and dividing the superimposed surface electromyogram signal by the number of channels to obtain a single channel surface muscle
  • the electrical signal starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyographic signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining the window sequence corresponding to each sliding moment Calculating an average amplitude of the myoelectric signals of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the average amplitude of the surface electromyogram signal of the window sequence is the surface electromyographic signal of the window sequence
  • the average value of the absolute values of the amplitudes, the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N consecutive windows corresponding to N-1 sliding moments before the sliding moment, and N is greater than or equal to A positive integer of 2.
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective surface electromyogram signal.
  • the start time of the effective surface EMG signal is added to the preset time to obtain the cut-off time of the effective surface EMG signal, and the surface EMG signals of the plurality of channels between the start time and the cut-off time are taken as effective surfaces.
  • the preset amplitude may be an average value of the absolute values of the amplitudes of the signals superimposed by the surface electromyogram signals of the plurality of channels.
  • Step 303 Determine a frequency of the effective surface electromyogram signal.
  • Determining the frequency of the effective surface EMG signal specifically: calculating the correlation coefficient between the effective surface EMG signal and the plurality of sine and cosine matrices respectively, wherein the sine and cosine matrix are composed of a sine function and a cosine function of the fundamental frequency and the frequency doubling, each The fundamental frequency of the sine cosine matrix is different; determining whether the maximum correlation coefficient between the effective surface EMG signal and the correlation coefficient of the plurality of sine and cosine matrix is greater than a preset correlation coefficient; if the maximum correlation coefficient is greater than the preset correlation coefficient, the maximum is The fundamental frequency of the sine and cosine matrix corresponding to the correlation coefficient is taken as the frequency of the effective surface EMG signal.
  • Step 304 extracting amplitude characteristics of the effective surface myoelectric signal.
  • Extracting the amplitude characteristics of the effective surface EMG signal specifically: first, the effective surface muscle
  • the surface electromyogram signals of each channel of the electrical signal are respectively subjected to sliding window processing to obtain a plurality of sliding windows, for example, the sliding window has a width of 100 ms and a sliding interval of 100 ms.
  • calculating the average amplitude of each sliding window of the surface electromyogram signal of each channel of the effective surface electromyogram signal wherein the average amplitude of each sliding window is the magnitude of the surface electromyogram signal of each sliding window
  • the average of the absolute values of the effective surface EMG signal is the average amplitude of each sliding window as an amplitude characteristic of the effective surface EMG signal.
  • the step 303 and the step 304 are not in the order of execution, and the step 304 may be performed first, and then the step 303 is performed.
  • Step 305 Determine a limb motion corresponding to the surface electromyogram signal of the plurality of channels according to the amplitude characteristic of the effective surface electromyogram signal and the frequency of the effective surface electromyogram signal.
  • a plurality of candidate limb actions corresponding to the surface electromyogram signals of the plurality of channels are determined according to the frequency of the effective surface electromyogram signals.
  • each frequency may correspond to multiple limb actions, such as a frequency of 1 Hz. It can correspond to the following three types of limb movements: “Bumping”, “OK Gesture” and “Elbow Elbow”. Then, when the frequency of the surface electromyogram signal is determined to be 1 Hz, the surface electromyogram signal corresponds to the above three alternative limb actions.
  • the amplitude characteristics of the effective surface EMG signal are matched with the amplitude characteristics of a plurality of candidate limb movements that are pre-trained to obtain a limb motion that matches the amplitude characteristic of the effective surface EMG signal, and the effective surface muscle
  • the amplitude characteristic of the electrical signal matches the limb motion as the limb motion corresponding to the surface electromyogram signal of the plurality of channels.
  • the amplitude feature of the alternative limb motion can be used as a template, and a linear discriminant analysis (LDA) classifier can be used to identify the limb motion corresponding to the surface electromyogram signal.
  • LDA linear discriminant analysis
  • the limb motion corresponding to the surface electromyogram signal is recognized according to the frequency and amplitude characteristics of the surface electromyogram signal, and the action based on the surface myoelectric signal can be improved.
  • the accuracy of recognition and can increase the type of surface EMG signal recognition, because the same limb movement can be regarded as different limb movements at different frequencies. Therefore, this embodiment has a significant improvement in the number of identification types.
  • FIG. 4 is an overall block diagram of a method for identifying a surface electromyogram signal according to Embodiment 4 of the present invention.
  • the identification method of the present embodiment divides the surface EMG signal identification process into two parts: based on amplitude. Feature recognition method and frequency-based recognition method.
  • the surface EMG signal should be trained first and trained. Practice the template.
  • the training process of the surface myoelectric signal specifically includes the following steps:
  • the surface EMG signals of multiple channels of various limb movements are acquired.
  • the surface EMG signals of multiple channels of each limb movement are pretreated.
  • a 50 Hz power frequency interference notch is applied to the collected multi-channel surface EMG signal, and a FIR filter is used for high-pass filtering to obtain a pre-processed surface EMG signal.
  • the amplitude characteristics of the surface electromyogram signals of the plurality of channels of each limb movement are extracted.
  • the user repeats the fist movement for 30 times, the time of each acquisition is [t1, t2, ..., t30], and the acquired surface electromyogram signal is composed of 8 channels of signals.
  • the acquired surface electromyogram signal is composed of 8 channels of signals.
  • it will be pre-processed in (T1, t1+300ms), (t2, t2+300ms),...(t30, t30+300ms) for a total of 30 time periods.
  • the signals of each channel of the surface EMG signal are slid according to the sliding window interval of 100ms, and the width of the sliding window is also 100ms.
  • the average value M(n) of the absolute values of the signals of the windows in the 30 time periods is calculated, and M (n)
  • the amplitude characteristics of all limb movements can be obtained in the same way.
  • the fourth step, the production of the training template is to establish the correspondence between each limb movement and the amplitude feature, and then the prepared training template is sent to the classifier, and the classifier performs the recognition of the limb motion according to the training template.
  • the above four steps are the training phase.
  • the identification phase the amplitude characteristics of the surface EMG signal are also extracted, and then the amplitude features are input into the classifier for identification.
  • the first step is to collect the surface EMG signals of the multiple channels; the second step is to perform the surface EMG signals of the multiple channels.
  • the third step is to determine the frequency of the effective surface EMG signal.
  • the CCA calculation is performed, that is, the correlation coefficient between the effective surface EMG signal and the multiple sine and cosine matrix is calculated, and the maximum correlation of the effective surface muscle edge signal is obtained.
  • Coefficient; the fifth step is to determine the frequency of the effective surface EMG signal.
  • the specific implementation manners of the first step to the fifth step are described in the first embodiment and the second embodiment, and are not described here. After determining the frequency of the effective surface EMG signal, the candidate limb motion is determined based on the frequency of the effective surface EMG signal.
  • the amplitude of the effective surface EMG signal is extracted.
  • the value feature inputs the amplitude feature of the effective surface EMG signal into the classifier, and the classifier uses the amplitude feature of the candidate limb motion as a template, according to the amplitude characteristics of the input effective surface EMG signal and the amplitude of the candidate limb motion.
  • the feature features are matched to obtain the action of the limb corresponding to the effective surface EMG signal.
  • the method of the embodiments of the present invention has a wide range of application scenarios: (1) as a start command of the wearable device: for example, by turning the wrist at a specific frequency multiple times to activate the myoelectric wristband to avoid misoperation; (2) driving, Use in non-stationary state such as running: for example, adjusting music volume, switching songs, answering calls, etc.; (3) Since the accuracy of recognition is high, it can be applied to trigger control commands.
  • disabled people control the direction and speed of the wheelchair through different rhythm gestures; (4) air mouse, for example, giving the rotation of different frequencies of the finger to the corresponding mouse operation; (5) simple game control: fast and slow fingers The two movements can respectively correspond to the acceleration and braking of the racing game; (6) the user's physical coordination and control ability can be tested during the rehabilitation treatment.
  • a limb motion that matches the amplitude characteristic of the effective surface EMG signal acts as a limb motion corresponding to the surface electromyogram signal of the plurality of channels.
  • the motion recognition device based on the surface electromyogram signal of the embodiment includes: an acquisition module 11 and a first The determination module 12, the second determination module 13, and the identification module 14 are determined.
  • the obtaining module 11 is configured to acquire surface electromyogram signals of the plurality of channels;
  • a first determining module 12 configured to determine an effective surface electromyogram signal according to a surface electromyogram signal of the plurality of channels
  • a second determining module 13 configured to determine a frequency of the effective surface electromyogram signal
  • the identification module 14 is configured to determine a limb motion corresponding to the surface electromyogram signal of the plurality of channels according to the frequency of the effective surface electromyogram signal.
  • the first determining module 12 is specifically configured to:
  • the average amplitude of the surface electromyogram signal of the window sequence is an average value of the absolute values of the amplitudes of the surface myoelectric signals in the window sequence
  • the window sequence corresponding to the sliding moment includes the window corresponding to the sliding moment and the N-1 windows corresponding to the N-1 sliding moments before the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T in each sliding moment is taken as the effective surface
  • the start time of the EMG signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time and the cutoff of the effective surface EMG signal are intercepted.
  • a surface electromyogram signal of the plurality of channels between times acts as the effective surface myoelectric signal.
  • the preset amplitude may be an average value of the absolute values of the amplitudes of the surface electromyogram signals after the superposition of the surface electromyogram signals of the plurality of channels.
  • the second determining module 13 is specifically configured to: first calculate a correlation coefficient of the effective surface electromyogram signal and a plurality of sine and cosine matrices, wherein the sine and cosine matrix are sinusoidal functions of a fundamental frequency and a frequency multiplication And a cosine function, the fundamental frequency of each sine and cosine matrix is different; and then determining whether a maximum correlation coefficient between the effective surface EMG signal and the plurality of sine and cosine matrix correlation coefficients is greater than a preset correlation coefficient; And the maximum correlation coefficient is greater than the preset correlation coefficient, and the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is used as the frequency of the effective surface electromyogram signal.
  • the device in the embodiment of the present invention may be used to implement the solution in the first embodiment and the second embodiment.
  • the specific implementation and technical effects are similar, and details are not described herein again.
  • FIG. 6 is a schematic structural diagram of a motion recognition device based on a surface electromyogram according to a sixth embodiment of the present invention.
  • the motion recognition device based on the surface electromyogram signal according to the embodiment includes: an acquisition module 21, A determination module 22, a second determination module 23, an extraction module 24 and an identification module 25.
  • the obtaining module 21 is configured to acquire surface electromyogram signals of the plurality of channels;
  • a first determining module 22 configured to determine an effective surface electromyogram signal according to a surface electromyogram signal of the plurality of channels
  • a second determining module 23 configured to determine a frequency of the effective surface electromyogram signal
  • An extraction module 24 configured to extract an amplitude feature of the effective surface myoelectric signal
  • the identification module 25 is configured to determine a limb motion corresponding to the surface electromyogram signal of the plurality of channels according to the amplitude feature of the effective surface electromyogram signal and the frequency of the effective surface electromyogram signal.
  • the first determining module 22 is specifically configured to:
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective surface
  • the start time of the EMG signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time and the cutoff of the effective surface EMG signal are intercepted.
  • a surface electromyogram signal of the plurality of channels between times acts as the effective surface myoelectric signal.
  • the preset amplitude may be an average value of the absolute values of the amplitudes of the surface electromyogram signals after the superposition of the surface electromyogram signals of the plurality of channels.
  • the second determining module 23 is specifically configured to: first calculate a correlation coefficient between the effective surface electromyogram signal and a plurality of sine and cosine matrices, wherein the sine and cosine matrix are sinusoidal functions of a fundamental frequency and a frequency doubling And a cosine function, the fundamental frequency of each sine and cosine matrix is different; and then determining whether a maximum correlation coefficient between the effective surface EMG signal and the plurality of sine and cosine matrix correlation coefficients is greater than a preset correlation coefficient; And the maximum correlation coefficient is greater than the preset correlation coefficient, and the fundamental frequency of the sine and cosine matrix corresponding to the maximum correlation coefficient is used as the frequency of the effective surface electromyogram signal.
  • the extraction module 24 is specifically configured to: perform sliding window processing on the surface electromyogram signals of each channel of the effective surface electromyogram signal; calculate surface electromyography of each channel of the effective surface electromyogram signal An average amplitude of each sliding window of the signal, wherein the average amplitude of each of the sliding windows is an average of absolute values of amplitudes of the surface electromyographic signals of each of the sliding windows, the effective surface
  • the average amplitude of each sliding window of the myoelectric signal is taken as the amplitude characteristic of the effective surface myoelectric signal.
  • the identification module 25 is specifically configured to: determine, according to the frequency of the effective surface electromyogram signal, a plurality of candidate limb motions corresponding to the surface electromyogram signals of the plurality of channels; and the effective surface electromyogram signal Amplitude features are matched to amplitude features previously trained to obtain the plurality of candidate limb motions to obtain a limb motion that matches the amplitude characteristics of the effective surface EMG signal, and the effective surface EMG signal
  • the limb feature matching limb motion acts as a limb motion corresponding to the surface electromyogram signal of the plurality of channels.
  • the device in this embodiment may be used to perform the technical solution provided in the third embodiment, and the specific implementation manners and technical effects are similar, and details are not described herein again.
  • FIG. 7 is a schematic structural diagram of a motion recognition device based on a surface electromyogram according to a seventh embodiment of the present invention.
  • the motion recognition device 300 based on the surface electromyogram signal includes: a processor 31, The memory 32 and the system bus 33, the processor 31 and the memory 32 are connected through the system bus 33 and complete communication with each other; the memory 32 is used to store computer execution instructions 321; and the processor 31 is configured to execute a computer to execute instructions 321, perform the method described below:
  • a limb motion corresponding to a surface electromyogram signal of the plurality of channels is determined according to a frequency of the effective surface electromyogram signal.
  • the processor 31 when determining the effective surface myoelectric signal according to the surface electromyogram signals of the plurality of channels, the processor 31 is specifically configured to:
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T in each sliding moment is taken as the effective surface
  • the start time of the EMG signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time and the cutoff of the effective surface EMG signal are intercepted.
  • a surface electromyogram signal of the plurality of channels between times acts as the effective surface myoelectric signal.
  • the preset amplitude may be an average value of the absolute values of the amplitudes of the surface electromyogram signals after the superposition of the surface electromyogram signals of the plurality of channels.
  • the processor 31 when determining the frequency of the effective surface electromyogram signal, is specifically configured to: separately calculate correlation coefficients of the effective surface electromyogram signal and the plurality of sine and cosine matrices, wherein the sine and cosine matrix Comprising a sine function of a fundamental frequency and a frequency doubling and a cosine function, the fundamental frequency of each sine and cosine matrix is different; determining whether a maximum correlation coefficient between the effective surface EMG signal and the correlation coefficient of the plurality of sine and cosine matrices is greater than a preset correlation coefficient; if the maximum correlation coefficient is greater than the preset correlation coefficient, a fundamental frequency of a sine and cosine matrix corresponding to the maximum correlation coefficient is used as a frequency of the effective surface electromyogram signal.
  • the device in this embodiment may be used to implement the technical solutions in the first embodiment and the second embodiment.
  • the specific implementation manners and technical effects are similar, and details are not described herein again.
  • the motion recognition device 400 based on the surface electromyogram signal of the present embodiment includes: a processor 41 and a memory. 42 and a system bus 43, the processor 41 and the memory 42 are connected through the system bus 43 and complete communication with each other; the memory 42 is configured to store a computer execution instruction 421; the processor 41 For executing the computer execution instructions 421 to perform the methods described below:
  • a limb motion corresponding to a surface electromyogram signal of the plurality of channels is determined based on an amplitude characteristic of the effective surface electromyogram signal and a frequency of the effective surface electromyogram signal.
  • the processor 41 when determining the effective surface myoelectric signal according to the surface electromyogram signals of the plurality of channels, the processor 41 is specifically configured to:
  • a single channel surface EMG signal is obtained by dividing the number of channels;
  • the surface electromyogram signal of the single channel Starting from the start time of the single-channel surface electromyogram signal, sliding the surface electromyogram signal of the single channel at each sliding moment to obtain a window corresponding to each sliding moment, and determining each sliding a window sequence corresponding to the moment, calculating an average amplitude of the surface electromyogram signal of the window sequence corresponding to each sliding moment, wherein each sliding moment differs by a sliding interval, and the surface EMG signal of the window sequence
  • the average amplitude is an average value of absolute values of amplitudes of surface electromyogram signals in the window sequence
  • the window sequence corresponding to the sliding moment includes a window corresponding to the sliding moment and N-1 times before the sliding moment N-1 windows corresponding to the sliding moment have a total of N consecutive windows, and N is a positive integer greater than or equal to 2;
  • the starting time of the window sequence corresponding to the sliding time T is taken as the effective surface
  • the start time of the EMG signal, the start time of the effective surface EMG signal is added to the preset time to obtain the cutoff time of the effective surface EMG signal, and the start time and the cutoff of the effective surface EMG signal are intercepted.
  • a surface electromyogram signal of the plurality of channels between times acts as the effective surface myoelectric signal.
  • the preset amplitude may be an average value of absolute values of amplitudes of surface electromyogram signals superposed by surface electromyogram signals of the plurality of channels.
  • the processor 41 when determining the frequency of the effective surface electromyogram signal, is specifically configured to: separately calculate correlation coefficients of the effective surface electromyogram signal and the plurality of sine and cosine matrices, wherein the sine and cosine matrix Comprising a sine function of a fundamental frequency and a frequency doubling and a cosine function, the fundamental frequency of each sine and cosine matrix is different; determining whether a maximum correlation coefficient between the effective surface EMG signal and the correlation coefficient of the plurality of sine and cosine matrices is greater than a preset correlation coefficient; if the maximum correlation coefficient is greater than the preset correlation coefficient, a fundamental frequency of a sine and cosine matrix corresponding to the maximum correlation coefficient is used as a frequency of the effective surface electromyogram signal.
  • the processor 41 when extracting the amplitude feature of the effective surface electromyogram signal, is specifically configured to: perform sliding window processing on the surface electromyogram signal of each channel of the effective surface electromyogram signal; An average amplitude of each sliding window of the surface electromyogram signal of each channel of the effective surface electromyogram signal, wherein the average amplitude of each of the sliding windows is an intramuscular signal of the inner surface of each of the sliding windows.
  • the average of the absolute values of the amplitudes, the average amplitude of each sliding window of the effective surface EMG signal is taken as the amplitude characteristic of the effective surface EMG signal.
  • the processor 41 determines a limb motion corresponding to the surface electromyogram signal of the plurality of channels, Specifically, determining, according to the frequency of the effective surface electromyogram signal, a plurality of candidate limb motions corresponding to the surface electromyogram signals of the plurality of channels; and obtaining the amplitude characteristics of the effective surface EMG signal and pre-training Matching the amplitude features of the plurality of candidate limb movements to obtain a limb motion that matches the amplitude characteristics of the effective surface EMG signal, and the limb motion that matches the amplitude characteristics of the effective surface EMG signal A limb motion corresponding to a surface electromyogram signal of the plurality of channels.
  • the device in this embodiment can be used to perform the technical solution in the third embodiment, and the specific implementation manners and technical effects are similar, and details are not described herein again.
  • the aforementioned program can be stored in a computer readable storage medium.
  • the program when executed, performs the steps including the foregoing method embodiments; and the foregoing storage medium includes various media that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Veterinary Medicine (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

一种基于表面肌电信号的动作识别方法，包括：获取多个通道的表面肌电信号（101,301）；根据多个通道的表面肌电信号确定有效表面肌电信号（102,302）；确定有效表面肌电信号的频率（103,303）；根据有效表面肌电信号的频率确定多个通道的表面肌电信号对应的肢体动作（104）。由于表面肌电信号的频率与信号强度等特征无关，因此，该方法能显著提高基于表面肌电信号的动作识别的准确率。而且，以频率作为识别特征，用户不需要进行大幅度的动作，给用户带来更好的体验。另外，这种基于表面肌电信号的动作识别方法，还可包括：提取有效表面肌电信号的幅值特征（304）；根据有效表面肌电信号的幅值特征和有效表面肌电信号的频率来确定多个通道的表面肌电信号对应的肢体动作（305）。这一方法不仅可以提高表面肌电信号识别准确率，而且能够增加肢体动作的识别种类。

Description

基于表面肌电信号的动作识别方法和设备 技术领域
本发明实施例涉及动作识别技术，尤其涉及一种基于表面肌电信号的动作识别方法和设备。
背景技术
表面肌电信号(Electromyography，简称EMG)是一种与神经肌肉活动相关的生物电信号，表面肌电信号能够反应肌肉的收缩模式以及收缩强度等信息，不同的肢体动作对应不同的表面肌电信号，通过分析表面肌电信号可以判别出该表面肌电信号对应的具体动作，因此，表面肌电信号被广泛应用于临床医学、运动医学、生物医学与康复工程等诸多领域。
现有技术中，以表面肌电信号的幅值特征作为识别参数进行肢体动作的识别，该方法通过将采集到的表面肌电信号取滑动加窗，得到每个滑动窗口的窗口序列，然后，计算每个窗口的窗口序列的幅度得到幅值特征，将该幅值特征与预先训练得到的各肢体动作对应的表面肌电信号幅值特征进行比较以确定该幅值特征对应的肢体动作。
现有技术的方法存在以下问题：由于用户的其他动作也会产生表面肌电信号，当用户进行其他运动(跑步、打字等)时，产生的干扰表面肌电信号与目标表面肌电信号会发生叠加，导致识别准确率低，另外，表面肌电信号的幅值的大小正比于用户肢体动作的幅度，为了达到较高的信噪比，用户需要进行较大的力度的肢体动作，在长时间操作的情况下，会使用户感到吃力。
发明内容
本发明实施例提供一种基于表面肌电信号的动作识别方法和设备，能够提高基于表面肌电信号的动作识别的准确率。
本发明第一方面提供一种基于表面肌电信号的动作识别方法，包括：
获取多个通道的表面肌电信号；
根据所述多个通道的表面肌电信号确定有效表面肌电信号；
确定所述有效表面肌电信号的频率；
根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
结合本发明第一方面，在本发明第一方面的第一种可能的实现方式中，所述根据所述多个通道的表面肌电信号确定有效表面肌电信号，包括：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
若所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
结合本发明第一方面的第一种可能的实现方式，在本发明第一方面的第二种可能的实现方式中，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
结合本发明第一方面以及本发明第一方面的第一种和第二种可能的实现方式，在本发明第一方面的第三种可能的实现方式中，所述确定所述有效表面肌电信号的频率，包括：
分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大 相关系数是否大于预设的相关系数；
若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
本发明第二方面提供一种基于表面肌电信号的动作识别方法，包括：
获取多个通道的表面肌电信号；
根据所述多个通道的表面肌电信号确定有效表面肌电信号；
确定所述有效表面肌电信号的频率；
提取所述有效表面肌电信号的幅值特征；
根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
结合本发明第二方面，在本发明第二方面的第一种可能的实现方式中，所述根据所述多个通道的表面肌电信号确定有效表面肌电信号，包括：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
当所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值时，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
结合本发明第二方面的第一种可能的实现方式，在本发明第二方面的第二种可能的实现方式中，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
结合本发明第二方面以及本发明第二方面的第一种和第二种可能的实现方式，在本发明第二方面的第三种可能的实现方式中，所述确定所述有效表面肌电信号的频率，包括：
分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；
若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
结合本发明第二方面的第三种可能的实现方式，在本发明第二方面的第四种可能的实现方式中，所述提取所述有效表面肌电信号的幅值特征，包括：
对所述有效表面肌电信号的每个通道的表面肌电信号分别进行滑动窗口处理；
计算所述有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，所述每个滑动窗口的平均幅值为所述每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将所述有效表面肌电信号的每个滑动窗口的平均幅值作为所述有效表面肌电信号的幅值特征。
结合本发明第二方面以及本发明第二方面的第一种和第二种可能的实现方式，在本发明第二方面的第五种可能的实现方式中，所述根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作，包括：
根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的多个备选肢体动作；
将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
本发明第三方面提供一种基于表面肌电信号的动作识别设备，包括：
获取模块，用于获取多个通道的表面肌电信号；
第一确定模块，用于根据所述多个通道的表面肌电信号确定有效表面肌 电信号；
第二确定模块，用于确定所述有效表面肌电信号的频率；
识别模块，用于根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
结合本发明第三方面，在本发明第三方面的第一种可能的实现方式中，所述第一确定模块具体用于：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
若所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
结合本发明第三方面的第一种可能的实现方式，在本发明第三方面的第二种可能的实现方式中，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
结合本发明第三方面以及第三方面的第一种和第二种可能的实现方式，在本发明第三方面的第三种可能的实现方式中，所述第二确定模块具体用于：
分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大 相关系数是否大于预设的相关系数；
若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
本发明第四方面提供一种基于表面肌电信号的动作识别设备，包括：
获取模块，用于获取多个通道的表面肌电信号；
第一确定模块，用于根据所述多个通道的表面肌电信号确定有效表面肌电信号；
第二确定模块，用于确定所述有效表面肌电信号的频率；
提取模块，用于提取所述有效表面肌电信号的幅值特征；
识别模块，用于根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
结合本发明第四方面，在本发明第四方面的第一种可能的实现方式中，所述第一确定模块具体用于：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
当所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值时，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
结合本发明第四方面的第一种可能的实现方式，在本发明第四方面的第二种可能的实现方式中，所述预设幅值为所述多个通道的表面肌电信号叠加 后的表面肌电信号的幅值的绝对值的平均值。
结合本发明第四方面以及第四方面的第一种和第二种可能的实现方式，在本发明第四方面的第三种可能的实现方式中，所述第二确定模块具体用于：
分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；
若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
结合本发明第四方面的第三种可能的实现方式，在本发明第四方面的第四种可能的实现方式中，所所述提取模块具体用于：
对所述有效表面肌电信号的每个通道的表面肌电信号分别进行滑动窗口处理；
计算所述有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，所述每个滑动窗口的平均幅值为所述每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将所述有效表面肌电信号的每个滑动窗口的平均幅值作为所述有效表面肌电信号的幅值特征。
结合本发明第四方面以及第四方面的第一种和第二种可能的实现方式，在本发明第四方面的第五种可能的实现方式中，所述识别模块具体用于：
根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的多个备选肢体动作；
将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
本发明第五方面提供一种基于表面肌电信号的动作识别设备，包括：
处理器、存储器和系统总线，所述处理器和所述存储器之间通过所述系统总线连接并完成相互间的通信；
所述存储器，用于存储计算机执行指令；
所述处理器，用于运行所述计算机执行指令，使所述动作识别设备执行 本发明第一方面以及本发明第一方面的第一种至第三种可能的实现方式提供的方法。
本发明第六方面提供一种基于表面肌电信号的动作识别设备，包括：
处理器、存储器和系统总线，所述处理器和所述存储器之间通过所述系统总线连接并完成相互间的通信；
所述存储器，用于存储计算机执行指令；
所述处理器，用于运行所述计算机执行指令，使所述动作识别设备执行本发明第二方面以及本发明第二方面的第一种至第五种可能的实现方式提供的方法。
本发明实施例提供的基于表面肌电信号的动作识别方法和设备，通过获取多个通道的表面肌电信号，根据多个通道的表面肌电信号确定有效表面肌电信号；然后，确定有效表面肌电信号的频率，根据有效表面肌电信号的频率确定多个通道的表面肌电信号对应的肢体动作。由于表面肌电信号的频率与信号强度等特征无关，因此，本实施例的方法能显著提高基于表面肌电信号的动作识别的准确率。而且以频率作为识别特征，用户不需要进行大幅度的动作，给用户带来更好的体验。另外，本发明实施例的方法，还可以将表面肌电信号的频率和幅值特征结合起来识别表面肌电信号对应的肢体动作，不仅可以提高表面肌电信号识别准确率，而且能够增加肢体动作的识别种类。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本发明实施例一提供的基于表面肌电信号的动作识别方法的流程图；
图2为本发明实施例二提供的一种有效表面肌电信号的频率确定方法；
图3为本发明实施例三提供的基于表面肌电信号的动作识别方法的流程图；
图4为本发明实施例四提供的基于表面肌电信号的动作识别方法的整体 框图；
图5为本发明实施例五提供的基于表面肌电信号的动作识别设备的结构示意图；
图6为本发明实施例六提供的基于表面肌电信号的动作识别设备的结构示意图；
图7为本发明实施例七提供的基于表面肌电信号的动作识别设备的结构示意图；
图8为本发明实施例八提供的基于表面肌电信号的动作识别设备的结构示意图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
图1为本发明实施例一提供的基于表面肌电信号的动作识别方法的流程图，本实施例的方法可以由基于表面肌电信号的动作识别设备执行，该基于表面肌电信号的动作识别设备具体可以为智能手机、平板电脑等终端设备。如图1所示，本实施例的方法可以包括以下步骤：
步骤101、获取多个通道的表面肌电信号。
基于表面肌电信号的动作识别设备获取多个通道的表面肌电信号具体为：接收采集装置发送的多个通道的表面肌电信号，通过将采集装置放置在要采集的肌肉群表面进行表面肌电信号的采集，由于本实施例的方法中，是根据表面肌电信号的频率确定对应的肢体动作，因此，在采集表面肌电信号时要求用户以特定的频率重复进行同一种肢体动作。该肢体动作不仅可以是人体上肢前臂的节律性手势，还可以是下肢腿部的节律性动作，甚至可以是脖子、腹部等躯体的节律性动作，这里节律性动作即按照一定频率重复的动作。
采集装置包括多个传感节点，每个传感节点采集的数据作为一个通道的 表面肌电信号，采集装置可以嵌入智能手表、手环等穿戴式设备中。在用户进行某一肢体动作时，可能会带动多个肌肉群运动，因此，需要对多个肌肉群都进行表面肌电信号采集，采集到的多个通道的信号综合反映用户的肢体动作。
步骤102、根据多个通道的表面肌电信号确定有效表面肌电信号。
由于表面肌电信号的幅度小、信噪比低，因此，在确定有效表面肌电信号之前需要对获取的多个通道的表面肌电信号进行预处理，对表面肌电信号进行预处理具体为：对表面肌电信号进行信号放大、工频滤波、高通滤波等处理。
采集装置采集到的表面肌电信号的最开始部分的信号可能有一些干扰信号，例如，采集装置工作后用户没有及时的做相应的肢体动作，此时用户虽然没有做肢体动作，采集装置仍然能够采集到一些微弱的表面肌电信号，这些微弱的表面肌电信号属于干扰信号，因此，需要祛除可能的干扰信号得到有效的表面肌电信号。
具体地，可以根据如下方法确定有效表面肌电信号：
第一步、将预处理后多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号。
第二步、从单通道的表面肌电信号的起始时间开始，在每个滑动时刻对单通道表面肌电信号进行滑动得到每个滑动时刻对应的窗口，确定每个滑动时刻对应的窗口序列，计算每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，每个滑动时刻之间相差一个滑动间隔，窗口序列的表面肌电信号平均幅值为窗口序列内表面肌电信号的幅值的绝对值的平均值，其中，滑动时刻对应的窗口序列包括滑动时刻对应的窗口以及滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数。
其中，单通道表面肌电信号的起始时间是指开始有通道信号的时间，在每个滑动时刻对单通道的表面肌电信号进行滑动，即在每个滑动时刻将单通道表面信号在时间上向后移动一个窗口，相邻两个滑动时刻之间相差的时间为一个滑动间隔。例如，窗口宽度为时间t，滑动间隔为t/2，t＝1ms(毫秒)，每个窗口序列包括四个窗口，共有50个窗口，窗口的编号为1、2、3……50，那么从单通道表面肌电信号的起始时间开始，第一个滑动时刻对应的窗口为4号窗口，第一个滑动时刻对应的窗口序列为0-4ms，共4个窗口，然后，在 第二个滑动时刻将单通道表面肌电信号向后滑动0.5ms，得到第二个窗口序列，第二个窗口序列为0.5-4.5ms，依次类推，得到每个滑动时刻对应的窗口序列。在计算第一个滑动时刻对应的窗口序列的表面肌电信号平均幅值时，首先，对该窗口序列内的单通道的表面肌电信号的值取绝对值，然后，将该窗口序列内的单通道的表面肌电信号的绝对值加起来取平均值，得到该窗口序列的表面肌电信号平均幅值。
第三步、若每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将滑动时刻T对应的窗口序列的起始时间作为有效表面肌电信号的起始时间，将有效表面肌电信号的起始时间加上预设时间得到有效表面肌电信号的截止时间，截取有效表面肌电信号的开始时间和截止时间之间的多个通道预处理后的表面肌电信号作为有效表面肌电信号。
在计算得到每个滑动时刻对应的窗口序列的表面肌电信号平均幅值时，判断该滑动时刻对应的窗口序列的表面肌电信号平均幅值是否小于预设幅值，若滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，即滑动时刻T对应的窗口序列的表面肌电信号平均幅值大于等于预设幅值，则将滑动时刻T对应的窗口序列的起始时间作为有效表面肌电信号的起始时间，有效表面肌电信号的起始时间加上预设时间得到有效表面肌电信号的截止时间。在得到有效表面肌电信号的开始时间和截止时间后，截取开始时间和截止时间之间多个通道的表面肌电信号作为有效表面肌电信号。其中，预设幅值可以为多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
步骤103、确定有效表面肌电信号的频率。
确定有效表面肌电信号的频率的方法有多种，以下列举几种常用的方法：第一种方式，计算有效表面肌电信号与多个正余弦矩阵的相关系数，假设肢体动作可能出现的频率为f1，…,fn，那么选取以f1,…,fn为基频的正余弦矩阵，分别将有效表面肌电信号与各正余弦矩阵进行典型相关分析(Canonical Correlation Analysis，简称CCA)运算，得到有效表面肌电信号与各正余弦矩阵的最大相关系数，如果最大相关系数大于预设的相关系数，则将最大相关系数对应的正余弦矩阵的基频作为有效表面肌电信号的频率。第二种方法，对有效表面信号进行快速傅里叶变换(Fast Fourier Ttransform，简称FFT)，根据变换结果确定有效表面肌电信号的频率分布，从而得到有效表面肌电信 号的频率。第三种方法，根据有效表面肌电信号两个过零点的时间间隔计算表面肌电信号的频率。
步骤104、根据有效表面肌电信号的频率确定多个通道的表面肌电信号对应的肢体动作。
预先需要建立表面肌电信号的频率与肢体动作的对应关系，例如，1HZ频率对应的肢体动作为“握拳”，2HZ频率对应的肢体动作为“胳膊肘弯曲”，3HZ频率对应的肢体动作为“ok手势”，那么当用户以1HZ的频率做握拳动作时，根据获取到的有效表面肌电信号的频率确定用户的肢体动作为握拳。
本发明实施例，通过获取多个通道的表面肌电信号，根据多个通道的表面肌电信号确定有效表面肌电信号；然后，确定有效表面肌电信号的频率，最后，根据表面肌电信号的频率确定多个通道的表面肌电信号对应的肢体动作。由于表面肌电信号的频率与信号强度等特征无关，因此，本实施例的方法能显著提高基于表面肌电信号的动作识别的准确率。而且以频率作为识别特征，用户不需要进行大幅度的动作，给用户带来更好的体验。
另外，本实施例的方法，具有很好的抗干扰性能，能有效抵抗无关动作所产生的肌电噪音干扰，用户可在跑步、开车、做家务等非静止条件下使用。并且不受使用过程中皮肤湿度变化、电极接触情况变化、肌肉疲劳程度等因素的影响，稳定性高。用户使用方便，无需事先采集用户训练数据，每次使用前也无需重新训练。
本发明实施例二中将对实施例一中步骤103进行详细说明，图2为本发明实施例二提供的一种有效表面肌电信号的频率确定方法，如图2所示，本实施例的方法包括以下步骤：
步骤201、分别计算有效表面肌电信号与多个正余弦矩阵的相关系数，其中，正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同。
假设肢体动作可能出现的频率为f1，…,fn，那么选取以f1,…,fn为基频的正余弦矩阵为对照。分别将有效表面肌电信号与这些正余弦矩阵进行CCA运算，获取有效表面肌电信号与各正余弦矩阵的相关系数。
记n通道的有效表面肌电信号为x＝(x1,x2,x3,...,xn)T，x1,x2,x3,...,xn分别代表每个通道的信号，正余弦矩阵为y＝[cos(2πf1t),sin(2πf1t),cos(4πf1t),sin(4πf1t)]T，f1是 基频。则CCA可以定义为如下问题：分别寻找向量wx和wy,使得x和y在向量wx和wx上的投影X＝xTwx和Y＝yTwy之间的相关值最大。也就是使得下式中的ρ最大：
Figure PCTCN2015073369-appb-000001
其中Cxy表示x,y的互相关矩阵，Cxx表示x的自相关矩阵，Cyy表示y的自相关矩阵。寻找向量wx和wy的方法如下：
第一步、通过拉格朗日算法，构建如下拉格朗日算子：
Figure PCTCN2015073369-appb-000002
第二步、将L分别对wx和wy求偏导，得到：
Cxywy-λxCxxwx＝0
Cyxwy-λyCyywy＝0，
从而得到公式1和公式2：
Figure PCTCN2015073369-appb-000003
λ＝λx＝λy,Cxx -1CxyCyy -1Cyxwx＝λ2wx   (2)
根据公式2可以求出wx，根据公式1可以求出wy，通过公式2可以看出根据公式2求wx的过程被转换为特征值分解问题，将wx和wy代入到ρ的定义中得到相关系数。
步骤202、判断有效表面肌电信号与多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数。
在计算出有效表面肌电信号与每个正余弦矩阵的相关系数后，找出最大相关系数，判断该最大相关系数是否大于预设的相关系数。
步骤203、若最大相关系数大于预设的相关系数，则将最大相关系数对应的正余弦矩阵的基频作为有效表面肌电信号的频率。
若最大相关系数大于预设的相关系数，说明有效表面肌电信号具备最大相关系数对应的正余弦矩阵的频率，将最大相关系数对应的正余弦矩阵的基频作为有效表面肌电信号的频率。
图3为本发明实施例三提供的基于表面肌电信号的动作识别方法的流程图，本实施例的方法与实施例一的区别在于：本实施例中，将结合表面肌电信号的频率和幅值特征确定表面肌电信号对应的肢体动作。
步骤301、获取多个通道的表面肌电信号。
步骤302、根据多个通道的表面肌电信号确定有效表面肌电信号。
根据多个通道的表面肌电信号确定有效表面肌电信号，具体为：将多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；从单通道的表面肌电信号的起始时间开始，在每个滑动时刻对单通道的表面肌电信号进行滑动得到每个滑动时刻对应的窗口，确定每个滑动时刻对应的窗口序列，计算每个滑动时刻对应的窗口序列的肌电信号平均幅值，其中，每个滑动时刻之间相差一个滑动间隔，窗口序列的表面肌电信号平均幅值为窗口序列内表面肌电信号的幅值的绝对值的平均值，滑动时刻对应的窗口序列包括滑动时刻对应的窗口以及滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数。当每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值时，将滑动时刻T对应的窗口序列的起始时间作为有效表面肌电信号的起始时间，将有效表面肌电信号的起始时间加上预设时间得到有效表面肌电信号的截止时间，截取开始时间和所述截止时间之间的多个通道的表面肌电信号作为有效表面肌电信号。其中，预设幅值可以为多个通道的表面肌电信号叠加后的信号的幅值的绝对值的平均值。
步骤303、确定有效表面肌电信号的频率。
确定有效表面肌电信号的频率，具体为：分别计算有效表面肌电信号与多个正余弦矩阵的相关系数，其中，正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；判断有效表面肌电信号与多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；若最大相关系数大于预设的相关系数，则将最大相关系数对应的正余弦矩阵的基频作为有效表面肌电信号的频率。
步骤301-303的具体实现方式可以参照实施例一和实施例二的相关描述，这里不再赘述。
步骤304、提取有效表面肌电信号的幅值特征。
提取有效表面肌电信号的幅值特征，具体可以为：首先，对有效表面肌 电信号的每个通道的表面肌电信号分别进行滑动窗口处理，得到多个滑动窗口，例如滑动窗口的宽度为100ms，滑动间隔为100ms。然后，计算有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，每个滑动窗口的平均幅值为每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将有效表面肌电信号的每个滑动窗口的平均幅值作为有效表面肌电信号的幅值特征。
需说明的时，本实施例中，步骤303和步骤304在执行时并没有先后顺序，也可以先执行步骤304，再执行步骤303。
步骤305、根据有效表面肌电信号的幅值特征和有效表面肌电信号的频率确定多个通道的表面肌电信号对应的肢体动作。
具体地，首先，根据有效表面肌电信号的频率确定多个通道的表面肌电信号对应的多个备选肢体动作，本实施例中，每个频率可以对应多个肢体动作，例如1HZ的频率可以分别对应以下三种肢体动作：“握拳”、“OK手势”和“胳膊肘弯曲”。那么，当确定表面肌电信号的频率为1HZ时，表面肌电信号对应上述三个备选肢体动作。然后，将有效表面肌电信号的幅值特征与预先训练得到多个备选肢体动作的幅值特征进行匹配，得到与有效表面肌电信号的幅值特征匹配的肢体动作，将与有效表面肌电信号的幅值特征匹配的肢体动作作为多个通道的表面肌电信号对应的肢体动作。可以将备选肢体动作的幅值特征作为模板，采用线性判别式分析(Linear Discriminant Analysis，简称LDA)分类器识别表面肌电信号对应的肢体动作。
本实施例的方法，通过获取表面肌电信号的频率和幅值特征，根据表面肌电信号的频率和幅值特征识别表面肌电信号对应的肢体动作，不仅可以提高基于表面肌电信号的动作识别的准确率，而且能够增加表面肌电信号识别种类，因为同一种肢体动作以不同频率做节律性运动可以视为不同的肢体动作，因此本实施例在识别种类数目上有明显的提升。
将表面肌电信号的频率和幅值特征结合确定表面肌电信号对应的肢体动作时，需要预先对表面肌电信号进行训练获取各种肢体动作的幅值特征。如图4所示，图4为本发明实施例四提供的表面肌电信号的识别方法的整体框图，本实施例的识别方法将表面肌电信号的识别过程分为两个部分：基于幅值特征识别方法和基于频率的识别方法。
在基于幅值特征的识别方法中，先要对表面肌电信号进行训练，得到训 练模板。表面肌电信号的训练过程具体包括以下步骤：
第一步，采集各种肢体动作的多个通道的表面肌电信号。
让用户将不同肢体动作(例如“握拳”和“OK”)重复做多遍，通过采集装置采集肢体动作关联的主要肌肉群的表面肌电信号，采集的表面肌电信号由多个通道组成，并记录每遍采集时间的时间。
第二步，对每种肢体动作的多个通道的表面肌电信号进行预处理。
例如，对采集到的多通道的表面肌电信号进行50HZ的工频干扰陷波，并使用FIR滤波器进行高通滤波，得到预处理后的表面肌电信号。
第三步，提取每种肢体动作的多个通道的表面肌电信号的幅值特征。
具体的提取方法可以参照实施例三种步骤304的描述，这里不再赘述。
举例来说，用户将握拳动作重复做30遍，每遍采集开始的时间为[t1，t2,…，t30]，采集到的表面肌电信号由8个通道的信号组成。在提取表面肌电信号的幅值特征时，依次在(t1,t1+300ms)，(t2,t2+300ms),…(t30,t30+300ms)共30个时间段内，将预处理后的表面肌电信号的各通道的信号按照100ms的滑动窗间隔进行滑动，滑动窗口的宽度也为100ms，计算这30个时间段内各窗口的信号的绝对值的平均值M(n)，将M(n)作为握拳肢体动作的幅值特征，按照同样的方法可以得到所有肢体动作的幅值特征。
第四步、训练模板的制作，即建立每种肢体动作与幅值特征的对应关系，然后，制作好的训练模板发送给分类器，由分类器根据训练模板进行肢体动作的识别。
上述四个步骤为训练阶段，在识别阶段，同样要提取表面肌电信号的幅值特征，然后，将幅值特征输入分类器进行识别。
在基于频率的识别方法中，不需要对表面肌电信号进行训练，在识别阶段，第一步，采集多个通道的表面肌电信号；第二步，对多个通道的表面肌电信号进行预处理；第三步，确定有效表面肌电信号的频率；第四步，进行CCA计算，即计算有效表面肌电信号与多个正余弦矩阵的相关系数，得到有效表面肌边信号的最大相关系数；第五步，确定有效表面肌电信号的频率。上述第一步至第五步的具体实现方式可操作实施例一和实施例二描述，这里不再赘述。在确定有效表面肌电信号的频率后，根据有效表面肌电信号的频率确定备选肢体动作。
本实施例中，在确定有效表面肌电信号后，提取有效表面肌电信号的幅 值特征，将有效表面肌电信号的幅值特征输入分类器，分类器将备选肢体动作的幅值特征作为模板，根据输入的有效表面肌电信号的幅值特征和备选肢体动作的幅值特征进行特征匹配，得到有效表面肌电信号对应的肢体动作。
本发明各实施例的方法具有广泛的应用场景：(1)作为穿戴设备的启动命令：例如通过多次、按特定频率转动手腕来启动肌电手环，避免误操作；(2)在开车、跑步等非静止状态下使用：例如调节音乐音量、切换歌曲、接听电话等；(3)由于识别的准确度高，因此可以应用于触发控制指令。比如：残疾人通过不同节律的手势来控制轮椅的方向和速度；(4)空中鼠标，例如将手指不同频率的转动赋予相应的鼠标的操作；(5)简单的游戏控制：手指的快，慢2个运动可以分别对应赛车游戏的加速和刹车；(6)用于康复治疗过程中对用户的身体协调和控制能力进行测试。
将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
图5为本发明实施例五提供的基于表面肌电信号的动作识别设备的结构示意图，如图5所示，本实施例的基于表面肌电信号的动作识别设备包括：获取模块11、第一确定模块12、第二确定模块13和识别模块14。
其中，获取模块11，用于获取多个通道的表面肌电信号；
第一确定模块12，用于根据所述多个通道的表面肌电信号确定有效表面肌电信号；
第二确定模块13，用于确定所述有效表面肌电信号的频率；
识别模块14，用于根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
可选地，第一确定模块12具体用于：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔， 所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
若所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。其中，预设幅值可以为多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
可选地，第二确定模块13具体用于：首先，分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；然后，判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
本发明实施例的设备可用于执行方法实施例一和实施例二的方案，具体实现方式和技术效果类似，这里不再赘述。
图6为本发明实施例六提供的基于表面肌电信号的动作识别设备的结构示意图，如图6所示，本实施例提供的基于表面肌电信号的动作识别设备包括：获取模块21、第一确定模块22、第二确定模块23、提取模块24和识别模块25。
其中，获取模块21，用于获取多个通道的表面肌电信号；
第一确定模块22，用于根据所述多个通道的表面肌电信号确定有效表面肌电信号；
第二确定模块23，用于确定所述有效表面肌电信号的频率；
提取模块24，用于提取所述有效表面肌电信号的幅值特征；
识别模块25，用于根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
可选地，第一确定模块22具体用于：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
当所述每个滑动时刻中的滑动时刻T对应的窗口序列的肌电信号平均幅值不小于预设幅值时，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。其中，预设幅值可以为多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
可选地，第二确定模块23具体用于：首先，分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；然后，判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
可选地，提取模块24具体用于：对所述有效表面肌电信号的每个通道的表面肌电信号分别进行滑动窗口处理；计算所述有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，所述每个滑动窗口的平均幅值为所述每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将所述有效表面肌电信号的每个滑动窗口的平均幅值作为所述有效表面肌电信号的幅值特征。
可选地，识别模块25具体用于：根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的多个备选肢体动作；将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
本实施例的设备可用于执行实施例三提供的技术方案，具体实现方式和技术效果类似，这里不再赘述。
图7为本发明实施例七提供的基于表面肌电信号的动作识别设备的结构示意图，如图7所示，本实施例的基于表面肌电信号的动作识别设备300，包括：处理器31、存储器32和系统总线33，处理器31和存储器32之间通过所述系统总线33连接并完成相互间的通信；存储器32，用于存储计算机执行指令321；处理器31，用于运行计算机执行指令321，执行如下所述的方法：
获取多个通道的表面肌电信号；
根据所述多个通道的表面肌电信号确定有效表面肌电信号；
确定所述有效表面肌电信号的频率；
根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
可选地，处理器31在根据所述多个通道的表面肌电信号确定有效表面肌电信号时，具体用于：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
若所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。其中，预设幅值可以为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
可选地，处理器31在确定所述有效表面肌电信号的频率时，具体用于：分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
本实施例的设备可用于执行实施例一和实施例二的技术方案，具体实现方式和技术效果类似，这里不再赘述。
图8为本发明实施例八提供的基于表面肌电信号的动作识别设备的结构示意图，如图8所示，本实施例的基于表面肌电信号的动作识别设备400包括：处理器41、存储器42和系统总线43，所述处理器41和所述存储器42之间通过所述系统总线43连接并完成相互间的通信；所述存储器42，用于存储计算机执行指令421；所述处理器41，用于运行所述计算机执行指令421执行如下所述的方法：
获取多个通道的表面肌电信号；
根据所述多个通道的表面肌电信号确定有效表面肌电信号；
确定所述有效表面肌电信号的频率；
提取所述有效表面肌电信号的幅值特征；
根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
可选地，处理器41在根据所述多个通道的表面肌电信号确定有效表面肌电信号时，具体用于：
将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号 除以通道数量得到单通道的表面肌电信号；
从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
当所述每个滑动时刻中的滑动时刻T对应的窗口序列的肌电信号平均幅值不小于预设幅值时，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。其中，所述预设幅值可以为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
可选地，处理器41在确定所述有效表面肌电信号的频率时，具体用于：分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
可选地，处理器41在提取所述有效表面肌电信号的幅值特征时，具体用于：对所述有效表面肌电信号的每个通道的表面肌电信号分别进行滑动窗口处理；计算所述有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，所述每个滑动窗口的平均幅值为所述每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将所述有效表面肌电信号的每个滑动窗口的平均幅值作为所述有效表面肌电信号的幅值特征。
可选地，处理器41在根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作时， 具体用于：根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的多个备选肢体动作；将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
本实施例的设备可用于执行实施例三的技术方案，具体实现方式和技术效果类似，这里不再赘述。
本领域普通技术人员可以理解：实现上述各方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成。前述的程序可以存储于一计算机可读取存储介质中。该程序在执行时，执行包括上述各方法实施例的步骤；而前述的存储介质包括：ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
最后应说明的是：以上各实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述各实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的范围。

Claims (22)

  1. 一种基于表面肌电信号的动作识别方法，其特征在于，包括：
    获取多个通道的表面肌电信号；
    根据所述多个通道的表面肌电信号确定有效表面肌电信号；
    确定所述有效表面肌电信号的频率；
    根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述多个通道的表面肌电信号确定有效表面肌电信号，包括：
    将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
    从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
    若所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
  3. 根据权利要求2所述的方法，其特征在于，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
  4. 根据权利要求1-3中任一项所述的方法，其特征在于，所述确定所述有效表面肌电信号的频率，包括：
    分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
    判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；
    若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
  5. 一种基于表面肌电信号的动作识别方法，其特征在于，包括：
    获取多个通道的表面肌电信号；
    根据所述多个通道的表面肌电信号确定有效表面肌电信号；
    确定所述有效表面肌电信号的频率；
    提取所述有效表面肌电信号的幅值特征；
    根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
  6. 根据权利要求5所述的方法，其特征在于，所述根据所述多个通道的表面肌电信号确定有效表面肌电信号，包括：
    将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
    从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
    当所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值时，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为 所述有效表面肌电信号。
  7. 根据权利要求6所述的方法，其特征在于，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
  8. 根据权利要求5-7任一项所述的方法，其特征在于，所述确定所述有效表面肌电信号的频率，包括：
    分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
    判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；
    若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
  9. 根据权利要求8所述的方法，其特征在于，所述提取所述有效表面肌电信号的幅值特征，包括：
    对所述有效表面肌电信号的每个通道的表面肌电信号分别进行滑动窗口处理；
    计算所述有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，所述每个滑动窗口的平均幅值为所述每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将所述有效表面肌电信号的每个滑动窗口的平均幅值作为所述有效表面肌电信号的幅值特征。
  10. 根据权利要求5-7中任一项所述的方法，其特征在于，所述根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作，包括：
    根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的多个备选肢体动作；
    将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
  11. 一种基于表面肌电信号的动作识别设备，其特征在于，包括：
    获取模块，用于获取多个通道的表面肌电信号；
    第一确定模块，用于根据所述多个通道的表面肌电信号确定有效表面肌电信号；
    第二确定模块，用于确定所述有效表面肌电信号的频率；
    识别模块，用于根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
  12. 根据权利要求11所述的设备，其特征在于，所述第一确定模块具体用于：
    将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
    从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
    若所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
  13. 根据权利要求12所述的设备，其特征在于，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
  14. 根据权利要求11-13中任一项所述的设备，其特征在于，所述第二确定模块具体用于：
    分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
    判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最 大相关系数是否大于预设的相关系数；
    若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
  15. 一种基于表面肌电信号的动作识别设备，其特征在于，包括：
    获取模块，用于获取多个通道的表面肌电信号；
    第一确定模块，用于根据所述多个通道的表面肌电信号确定有效表面肌电信号；
    第二确定模块，用于确定所述有效表面肌电信号的频率；
    提取模块，用于提取所述有效表面肌电信号的幅值特征；
    识别模块，用于根据所述有效表面肌电信号的幅值特征和所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的肢体动作。
  16. 根据权利要求15所述的设备，其特征在于，所述第一确定模块具体用于：
    将所述多个通道的表面肌电信号叠加在一起，对叠加后的表面肌电信号除以通道数量得到单通道的表面肌电信号；
    从所述单通道的表面肌电信号的起始时间开始，在每个滑动时刻对所述单通道的表面肌电信号进行滑动得到所述每个滑动时刻对应的窗口，确定所述每个滑动时刻对应的窗口序列，计算所述每个滑动时刻对应的窗口序列的表面肌电信号平均幅值，其中，所述每个滑动时刻之间相差一个滑动间隔，所述窗口序列的表面肌电信号平均幅值为所述窗口序列内表面肌电信号的幅值的绝对值的平均值，所述滑动时刻对应的窗口序列包括所述滑动时刻对应的窗口以及所述滑动时刻之前的N-1个滑动时刻对应的N-1个窗口共N个连续的窗口，N为大于等于2的正整数；
    当所述每个滑动时刻中的滑动时刻T对应的窗口序列的表面肌电信号平均幅值不小于预设幅值时，将所述滑动时刻T对应的窗口序列的起始时间作为所述有效表面肌电信号的起始时间，将所述有效表面肌电信号的起始时间加上预设时间得到所述有效表面肌电信号的截止时间，截取所述有效表面肌电信号的开始时间和截止时间之间的所述多个通道的表面肌电信号作为所述有效表面肌电信号。
  17. 根据权利要求16所述的设备，其特征在于，所述预设幅值为所述多个通道的表面肌电信号叠加后的表面肌电信号的幅值的绝对值的平均值。
  18. 根据权利要求15-17中任一项所述的设备，其特征在于，所述第二确定模块具体用于：
    分别计算所述有效表面肌电信号与多个正余弦矩阵的相关系数，其中，所述正余弦矩阵由基频和倍频的正弦函数与余弦函数组成，每个正余弦矩阵的基频不同；
    判断所述有效表面肌电信号与所述多个正余弦矩阵的相关系数中的最大相关系数是否大于预设的相关系数；
    若所述最大相关系数大于所述预设的相关系数，则将所述最大相关系数对应的正余弦矩阵的基频作为所述有效表面肌电信号的频率。
  19. 根据权利要求18所述的设备，其特征在于，所述提取模块具体用于：
    对所述有效表面肌电信号的每个通道的表面肌电信号分别进行滑动窗口处理；
    计算所述有效表面肌电信号的每个通道的表面肌电信号的每个滑动窗口的平均幅值，其中，所述每个滑动窗口的平均幅值为所述每个滑动窗口内表面肌电信号的幅值的绝对值的平均值，将所述有效表面肌电信号的每个滑动窗口的平均幅值作为所述有效表面肌电信号的幅值特征。
  20. 根据权利要求15-17中任一项所述的设备，其特征在于，所述识别模块具体用于：
    根据所述有效表面肌电信号的频率确定所述多个通道的表面肌电信号对应的多个备选肢体动作；
    将所述有效表面肌电信号的幅值特征与预先训练得到所述多个备选肢体动作的幅值特征进行匹配，得到与所述有效表面肌电信号的幅值特征匹配的肢体动作，将与所述有效表面肌电信号的幅值特征匹配的肢体动作作为所述多个通道的表面肌电信号对应的肢体动作。
  21. 一种基于表面肌电信号的动作识别设备，其特征在于，包括：
    处理器、存储器和系统总线，所述处理器和所述存储器之间通过所述系统总线连接并完成相互间的通信；
    所述存储器，用于存储计算机执行指令；
    所述处理器，用于运行所述计算机执行指令，使所述动作识别设备执行如权利要求1至4任一所述的方法。
  22. 一种基于表面肌电信号的动作识别设备，其特征在于，包括：
    处理器、存储器和系统总线，所述处理器和所述存储器之间通过所述系统总线连接并完成相互间的通信；
    所述存储器，用于存储计算机执行指令；
    所述处理器，用于运行所述计算机执行指令，使所述动作识别设备执行如权利要求5至10任一所述的方法。
PCT/CN2015/073369 2014-08-13 2015-02-27 基于表面肌电信号的动作识别方法和设备 WO2016023349A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP15832528.2A EP3167799A4 (en) 2014-08-13 2015-02-27 Movement recognition method and apparatus based on surface electromyographic signal
US15/428,230 US20170143226A1 (en) 2014-08-13 2017-02-09 Action Recognition Method and Device Based on Surface Electromyography Signal

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410396422.5 2014-08-13
CN201410396422.5A CN105326500B (zh) 2014-08-13 2014-08-13 基于表面肌电信号的动作识别方法和设备

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US15/428,230 Continuation US20170143226A1 (en) 2014-08-13 2017-02-09 Action Recognition Method and Device Based on Surface Electromyography Signal

Publications (1)

Publication Number Publication Date
WO2016023349A1 true WO2016023349A1 (zh) 2016-02-18

Family

ID=55277427

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/073369 WO2016023349A1 (zh) 2014-08-13 2015-02-27 基于表面肌电信号的动作识别方法和设备

Country Status (4)

Country Link
US (1) US20170143226A1 (zh)
EP (1) EP3167799A4 (zh)
CN (1) CN105326500B (zh)
WO (1) WO2016023349A1 (zh)

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106108897A (zh) * 2016-07-20 2016-11-16 西安中科比奇创新科技有限责任公司 一种基于经验模式分解的肌电信号滤波方法
CN107126303A (zh) * 2017-02-15 2017-09-05 上海术理智能科技有限公司 一种基于手机app的上下肢运动辅助方法
JP6590416B2 (ja) * 2017-07-10 2019-10-16 トヨタ自動車株式会社 リハビリ評価装置、リハビリ評価装置の作動方法およびリハビリ評価プログラム
CN108108016B (zh) * 2017-12-07 2020-07-14 浙江大学 手势感知器
CN108268844A (zh) * 2018-01-17 2018-07-10 上海术理智能科技有限公司 基于表面肌电信号的动作模式识别方法及装置
CN108403108B (zh) * 2018-02-11 2021-03-16 宁波工程学院 基于波形优化的阵列式表面肌电信号分解方法
CN109062401B (zh) * 2018-07-11 2020-12-15 北京理工大学 一种基于肌电信号的实时手势识别系统
CN109200469B (zh) * 2018-10-24 2020-05-05 山东大学 一种增强人工耳蜗音调感知的优化编码方法及系统
CN109512424B (zh) * 2018-11-16 2021-07-13 福州大学 一种高密度或多通道肌电信号的肌肉激活起点检测方法
CN109805923A (zh) * 2019-01-29 2019-05-28 北京京东方光电科技有限公司 可穿戴设备、信号处理方法及装置
CN109948465B (zh) * 2019-02-28 2020-11-13 浙江工业大学 一种基于多目标的表面肌电信号分类方法
CN110399846A (zh) * 2019-07-03 2019-11-01 北京航空航天大学 一种基于多通道肌电信号相关性的手势识别方法
CN113625882B (zh) * 2021-10-12 2022-06-14 四川大学 一种基于稀疏多通道相关性特征的肌电手势识别方法
CN113986017B (zh) * 2021-12-27 2022-05-17 深圳市心流科技有限公司 一种肌电手势模板的生成方法、装置及存储介质
CN115034273B (zh) * 2021-12-27 2023-09-01 驻马店市中心医院 一种基于模式识别的肌电生物反馈设备及系统
CN115299950A (zh) * 2022-08-12 2022-11-08 歌尔股份有限公司 肌电信号采集电路、可穿戴设备及控制方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070276280A1 (en) * 2003-11-19 2007-11-29 Urban Blomberg Method And Apparatus For Determining An Emg Signal
CN101995947A (zh) * 2009-08-21 2011-03-30 香港中文大学 通过网络感知用户动作的系统和方法
CN103495260A (zh) * 2013-06-04 2014-01-08 东南大学 一种基于肌电信号和传感器信号实现精细实时运动的控制方法
CN103654774A (zh) * 2014-01-02 2014-03-26 北京思睿博创科技有限公司 穿戴式移动手环

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8170656B2 (en) * 2008-06-26 2012-05-01 Microsoft Corporation Wearable electromyography-based controllers for human-computer interface
CN101587546B (zh) * 2009-06-12 2011-05-11 浙江大学 基于支持向量多元分类的下肢动作模式的识别方法
US8738122B2 (en) * 2009-08-21 2014-05-27 The Chinese University Of Hong Kong Systems and methods for reproducing body motions via networks
CN101930285B (zh) * 2009-11-23 2012-07-18 上海交通大学 基于表面肌电信号的手写识别方法
CN101874758B (zh) * 2010-05-07 2012-02-22 中国科学院深圳先进技术研究院 辅助运动系统及其控制方法
CN102139139A (zh) * 2011-01-13 2011-08-03 中国医学科学院生物医学工程研究所 一种肌电反馈控制的电刺激装置及其控制方法
CN102426651B (zh) * 2011-08-25 2014-04-16 武汉理工大学 人体前臂表面肌电信号采集及模式识别系统
CN102961203B (zh) * 2012-12-10 2015-04-22 杭州电子科技大学 基于emd样本熵的表面肌电信号识别方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070276280A1 (en) * 2003-11-19 2007-11-29 Urban Blomberg Method And Apparatus For Determining An Emg Signal
CN101995947A (zh) * 2009-08-21 2011-03-30 香港中文大学 通过网络感知用户动作的系统和方法
CN103495260A (zh) * 2013-06-04 2014-01-08 东南大学 一种基于肌电信号和传感器信号实现精细实时运动的控制方法
CN103654774A (zh) * 2014-01-02 2014-03-26 北京思睿博创科技有限公司 穿戴式移动手环

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3167799A4 *

Also Published As

Publication number Publication date
CN105326500A (zh) 2016-02-17
CN105326500B (zh) 2018-02-09
EP3167799A1 (en) 2017-05-17
EP3167799A4 (en) 2017-08-09
US20170143226A1 (en) 2017-05-25

Similar Documents

Publication Publication Date Title
WO2016023349A1 (zh) 基于表面肌电信号的动作识别方法和设备
EP2989967B1 (en) User registration method and corresponding user authentication apparatus based on electrocardiogram (ecg) signal
Li et al. Automatic recognition of sign language subwords based on portable accelerometer and EMG sensors
Fang et al. A multichannel surface EMG system for hand motion recognition
Saponas et al. Making muscle-computer interfaces more practical
Jiang et al. Open access dataset, toolbox and benchmark processing results of high-density surface electromyogram recordings
Mazilu et al. Feature learning for detection and prediction of freezing of gait in Parkinson’s disease
Nazarpour et al. Application of higher order statistics to surface electromyogram signal classification
Kim et al. A Study on EMG-based Biometrics.
Siddiqui et al. A wearable hand gesture recognition device based on acoustic measurements at wrist
Neacsu et al. Automatic EMG-based hand gesture recognition system using time-domain descriptors and fully-connected neural networks
Falzon et al. Complex-valued spatial filters for SSVEP-based BCIs with phase coding
CN108721048A (zh) 康复训练控制方法、计算机可读存储介质和终端
Ding et al. Motion intent recognition of individual fingers based on mechanomyogram
CN107273827B (zh) 一种心电信号r波检测方法及装置
Benatti et al. Towards EMG control interface for smart garments
Phinyomark et al. The relationship between anthropometric variables and features of electromyography signal for human–computer interface
Rivera et al. A power wheelchair controlled using hand gestures, a single sEMG sensor, and guided under-determined source signal separation
CN105686827B (zh) 一种基于微控制器的肌电信号处理及特征提取方法
CN109814716B (zh) 一种基于动态表面肌电信号的运动意图解码方法
Abdel-Hadi et al. Classification of motor imagery tasks with LS-SVM in EEG-based self-paced BCI
Anam et al. Index Finger Motion Recognition using self-advise support vector machine
Yu et al. Classification of nine directions using the maximum likelihood estimation based on electromyogram of both forearms
CN107684501B (zh) 一种基于表面肌电的肘关节动作连续辨识方法
CN111973184B (zh) 一种针对非理想sEMG信号的模型训练数据优化方法

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: 15832528

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

REEP Request for entry into the european phase

Ref document number: 2015832528

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 2015832528

Country of ref document: EP