KR101307515B1 - Apparatus for sensing bio-signal and method thereof - Google Patents

Abstract

The apparatus for recognizing a biological signal may include a plurality of sensor units attached to a living body, respectively, for sensing a EMG signal, a signal extractor for selectively extracting a part of a plurality of EMG signals sensed from the plurality of sensor units, and the extracted And a biometric state recognition unit configured to recognize at least one of a movement pattern of the living body and muscle fatigue based on an EMG signal. Accordingly, the pattern of the user's biological movement may be recognized or the degree of muscle fatigue according to the movement pattern may be determined.

Description

Apparatus for sensing bio-signal and method

The present invention relates to a biosignal recognition apparatus and a method thereof, and more particularly, to an apparatus and method for recognizing an EMG signal.

The EMG signal is one of biological signals generated in the specific muscle when the specific muscle contracts, and may be measured without impacting the user by the electronic device. The measured EMG signal is interpreted by various methods, thereby enabling the implementation of an interface between a human and an electronic device.

Conventional devices for measuring EMG signals have a limited number of channels, which makes it difficult to accurately determine the movement of muscles. The reason for the limited number of channels is that as the number of channels increases, the time and memory required to interpret the measured EMG signals increase rapidly.

As the time required to interpret the EMG signal increases, the real-time operation of the EMG signal measuring device becomes difficult, and there is a problem that the price of the EMG signal device increases as the required memory increases.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and an object of the present invention is to provide an apparatus and method for recognizing a biological signal using only EMG signals of some EMG signals collected by a multi-sensor.

In order to solve the above problems, a biosignal recognition apparatus according to an embodiment of the present invention, a plurality of sensor units respectively attached to the living body to sense the EMG signal (sensing), respectively sensed from the plurality of sensor units And a signal extractor configured to selectively extract some of a plurality of EMG signals and a biometric state recognizer configured to recognize at least one of a movement pattern of the living body and muscle fatigue based on the extracted EMG signals.

The apparatus may further include a storage unit configured to store an EMG signal threshold value for determining the validity of the EMG signal, wherein the signal extracting unit compares the EMG signal threshold value extracted from the storage unit with the EMG signal threshold. Only EMG signals above the value can be extracted.

The apparatus may further include an analog-to-digital converter (ADC) unit for performing an analog-to-digital conversion on the EMG signal extracted from the signal extractor to generate a digital value for the EMG signal.

The apparatus may further include a storage unit configured to store a feature map in which a region corresponding to at least one of the movement pattern or the muscle fatigue level is represented by a digital value, wherein the biometric recognition unit is configured to extract the digital value and the feature map extracted from the storage unit. At least one of the movement pattern and the muscle fatigue may be recognized based on the movement pattern.

The storage unit may further store time-series data corresponding to the digital value for a predetermined time, and the digital value used by the biometric recognition unit to recognize at least one of the movement pattern or the muscle fatigue degree is an absolute difference. It may be calculated using at least one of a standard deviation, an absolute integral average, a cumulative integral, and a zero crossing count.

The apparatus may further include a storage unit configured to store a corresponding relationship between the EMG signal and the movement pattern or the muscle fatigue, wherein the signal extractor may include the EMG signal received from the sensor unit and the EMG signal extracted from the storage unit, and Based on the correspondence relationship between the movement pattern or the muscle fatigue, one or more EMG signals may be extracted.

The signal extractor may extract a valid EMG signal from the plurality of EMG signals using principal component analysis (PCA) or independent component analysis (ICA).

The biometric state recognition unit uses at least one of a k-nearest neighbor method, a fuzzy c-means method, a quadratic discriminant analysis (QDA), and a linear discriminant analysis (LDA) algorithm. The pattern may be recognized using the clustered feature maps.

The apparatus may further include an output unit configured to output the movement pattern or the muscle fatigue information received from the biometric state recognition unit to the outside using a wired or wireless communication device.

The user interface unit receives an external signal for selecting some sensor units from the plurality of sensor units, a power unit for supplying power to the biosignal recognition apparatus, and the external signal, and receives power from the power unit. The apparatus may further include a switching unit configured to supply power to the sensor unit, wherein the signal extracting unit may extract an EMG signal from the sensor unit supplied with power from the switching unit.

The apparatus may further include an ADC unit receiving the EMG signal from the plurality of sensor units, converting an analog value into a digital value and transmitting the analog value to the signal extracting unit. The differential standard deviation may be used to extract some digital values from the plurality of digital values.

And a storage unit storing a threshold range for determining the validity of the digital value, generating a digital frequency value by performing a fast fourier transformation (FFT) on the digital EMG signal output from the ADC unit. The extractor may extract the digital frequency value included in the threshold range extracted from the storage unit.

Meanwhile, in a biosignal recognition method using a plurality of sensors attached to a living body, sensing a plurality of EMG signals of a living body for a predetermined time using the plurality of sensors, the sensed plurality Selectively extracting some of the EMG signals from the dog, and recognizing at least one of a movement pattern or muscle fatigue of the living body based on the extracted EMG signals.

In the extracting of the plurality of EMG signals, the EMG signal threshold value for determining the validity of the pre-stored EMG signal may be compared with the EMG signal, and the EMG-related EMG signal is greater than the EMG signal threshold value. Only signals can be extracted.

The method may further include generating a digital value of the EMG signal by performing an analog-digital conversion on the extracted EMG signal.

The digital value may be calculated using at least one of absolute difference standard deviation, absolute integral average value, cumulative integral value, and zero crossing count.

The recognizing of at least one of the movement pattern or the muscle fatigue degree based on the extracted partial EMG signal may include: a feature map in which a digital value and a region corresponding to the previously stored movement pattern or muscle fatigue degree are displayed as a digital value; Based on the movement pattern or the muscle fatigue.

The feature map may include at least one of a k-nearest neighbor method, a fuzzy c-means method, a quadratic discriminant analysis (QDA), and a linear discriminanat analysis (LDA) algorithm. Can be clustered.

In addition, selectively extracting a part of the sensed plurality of EMG signals, based on the corresponding relationship between the pre-stored EMG signal and the movement pattern of the living body, EMG corresponding to the stored pattern of the EMG signals collected Only signals can be extracted.

The method may further include outputting the pattern using a wired or wireless communication device.

In addition, selectively extracting a part of the sensed plurality of EMG signals may include effective EMG of the plurality of EMG signals using principal component analysis (PCA) or independent component analysis (ICA). The signal can be extracted.

And receiving an external signal for selecting some of the plurality of sensors and supplying power to the corresponding some sensor based on the external signal, wherein the plurality of sensed EMG signals are included. Selectively extracting a part may extract the EMG signal from the powered sensor.

Meanwhile, in a biosignal recognition method using a plurality of sensors attached to a living body, sensing a plurality of EMG signals of a living body for a predetermined time using the plurality of sensors, wherein the plurality of signals Generating a plurality of digital frequency values, selectively extracting a portion of the plurality of digital frequency values, and recognizing at least one of a movement pattern or muscle fatigue of the living body based on the extracted digital frequency values Steps.

The digital frequency value is generated by an FFT, and selectively extracting a part of the plurality of digital frequency values comprises selecting a digital frequency value within a threshold range for determining the validity of the digital frequency value. Can be extracted selectively.

According to various embodiments of the present disclosure, only the EMG signals of some of the EMG signals collected by the multi-sensor may be used to drastically reduce the biosignal recognition processing time and memory. By using signals extracted from various muscles, it is possible not only to accurately find information on the state and direction of muscle movement, but also to determine the degree of muscle fatigue.

1 is a conceptual diagram illustrating the extraction of some EMG signals by collecting EMG signals from a multi-sensor;
2 is a block diagram illustrating an apparatus for recognizing a physiological signal according to an embodiment of the present invention;
3A to 3D are block diagrams illustrating a process of extracting a part of a plurality of EMG signals by a signal extractor according to embodiments of the present disclosure;
3E is an exemplary view of implementing a user interface unit according to an embodiment of the present invention;
4A to 4D are graphs illustrating a process of forming and clustering feature maps;
5 is a block diagram for explaining a biosignal recognition apparatus according to another embodiment of the present invention;
6 is a flowchart illustrating a biosignal recognition method according to an embodiment of the present invention;
7 is a flowchart for explaining another biosignal recognition method of the present invention.

Hereinafter, various embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings.

1 is a conceptual diagram illustrating the extraction of some EMG signals by collecting EMG signals from a multi-sensor.

As illustrated in FIG. 1, the biosignal recognition apparatus may include a plurality of sensors 100, a signal extractor 200, and a biometric state recognizer 400.

The biosignal recognition apparatus according to an embodiment of the present invention may be attached to a body part such as a user's hand and wrist. Compared to a conventional biosignal recognition apparatus including electrodes disposed spaced apart from a body, one embodiment of the present invention may be attached to include a part of the body as shown in FIG. Although FIG. 1 is shown as a biosignal recognition device attached to a user's hand, various embodiments of the present invention may be applied to any part of the body. In addition, the biosignal recognition apparatus may be detachable to the body to facilitate use. In addition, the biosignal recognition apparatus may be flexibly manufactured so that there is no inconvenience when the user attaches to the body.

Each of the plurality of sensors 100 may include a means for collecting an EMG signal of a user. Sensor unit 100 according to an embodiment of the present invention may be implemented in the form of a plurality of electrodes. The number of electrodes can be adjusted to suit the purpose used. In addition, the individual electrode may be manufactured in a dry type to be semi-permanent use and may be composed of silver.

As shown in FIG. 1, the sensor units 100-1, 100-2, and 100-n implemented in the form of a plurality of electrodes sense an EMG signal generated from muscles attached to and attached to various muscles of a body ( Sensing). Since the contents of the sensing of the EMG signal by the electrode are well known technologies, a detailed description thereof will be omitted.

The EMG signals sensed and collected by the sensor units 100-1, 100-2, and 100-n are expressed in the form of analog signals. In FIG. 1, n EMG signals sensed and collected by the sensor units 100-1, 100-2, and 100-n are displayed in the form of respective channels.

That is, ch 1 may mean an analog EMG signal sensed and collected by the sensor unit 100-1, and ch n may mean an analog EMG signal sensed and collected by the sensor unit 100-n. . However, although each channel is shown in a combined state in FIG. 1, this does not mean that the signals corresponding to the channels are integrated and operated, respectively, and are merely for convenience of description.

The n EMG signals sensed and collected by the sensor units 100-1, 100-2, and 100-n are transmitted to the signal extractor 200.

The signal extractor 200 may receive n EMG signals from the sensor units 100-1, 100-2, and 100-n to extract only a portion of the n signals.

When the user moves the body, not all the muscles included in the moving body are relaxed or contracted together. Therefore, the EMG signal may not be detected in all of the plurality of sensor units 100-1, 100-2, and 100-n. Can be.

In addition, the EMG signal collected from the forearm muscle is used as useful data in the state of bending the wrist inward, for example, and the EMG signal collected from the finger is not useful.

Collecting n EMG signals as described above does not mean that all EMG signals are useful for grasping the user's movements. Will be increased.

The signal extractor 200 may reduce the working time and the memory used by extracting only the EMG signal suitable for detecting the user's movement and transmitting the signal to the next step. The configuration of extracting only the EMG signal suitable for detecting the user's movement by the signal extractor 200 will be described later in more detail.

The biometric state recognition unit 400 may recognize a movement pattern of the user or determine a degree of muscle fatigue with respect to the user's muscle by using some EMG signals received from the signal extractor 200. The configuration in which the biometric state recognition unit 400 recognizes a user's movement pattern and muscle fatigue will be described in more detail later.

2 is a block diagram illustrating a biosignal recognition apparatus according to an exemplary embodiment.

As shown in FIG. 2, the apparatus for recognizing a biosignal includes a sensor unit 100, a signal extractor 200, an ADC unit 300, a biometric state recognition unit 400, a storage unit 500, and an output unit 600. It may include.

In the embodiment related to FIG. 2, the sensor unit 100 may include six electrodes. The sensor unit 100 senses and collects EMG signals from six electrodes for a predetermined time. This is represented by reference numeral 10.

The six EMG signals sensed and collected by the sensor unit 100 are transmitted to the signal extractor 200. As described above, the signal extractor 200 extracts an EMG signal suitable for detecting a user's movement. In the present embodiment, the signal extractor 200 includes three sensor units 100-1 and 100-. 3, 100-5) may be used to extract the collected EMG signal. The extracted EMG signal is represented by reference numeral 20.

The extracted EMG signal 20 may be transmitted from the signal extractor 200 to the ADC 300. The ADC unit 300 may perform analog-to-digital conversion on the extracted EMG signal 20 in the form of an analog signal. The ADC unit 300 may convert the EMG signal 20 extracted in the form of an analog signal into a digital value 30. This is represented by reference numeral 30. The converted digital value 30 is transmitted to the biometric state recognition unit 400.

The biometric state recognition unit 400 may recognize the pattern of the user's movement or determine the degree of muscle fatigue of the user using the digital value 30 received from the ADC unit 300. The biometric state recognition unit 400 may store the digital value 30 received in the storage unit 500 in a time series. During the preset time, when the EMG signal is collected by the sensor unit 100, the biometric state recognition unit 400 extracts time-series digital values from the storage unit 500, and the time-series digital values are absolute difference standard for each channel. It is calculated as at least one of deviation, absolute integral value, cumulative integral value, and zero crossing count. The absolute difference standard deviation, the absolute integral average value, the cumulative integral value, and the zero crossing count may be calculated through Equations 1 to 4.

First, the absolute difference standard deviation may be calculated from Equation 1 below.

Equation 1 is an equation for absolute differential standard deviation, where X is an measured EMG signal, i is an order of acquired signals, and N is the number of signals obtained.

The absolute integrated average value IAV can be calculated from Equation 2 below.

Equation 2 is an expression for the absolute integral mean value, where X is the measured EMG signal, i is the order of the acquired signals, and N is the number of the acquired signals.

The cumulative integral value IEMG may be calculated from Equation 3 below.

Equation 3 is an expression for the cumulative integral value, where X is the measured EMG signal, i is the order of the acquired signals, and N is the number of the acquired signals.

The equation for the zero crossing frequency (ZCR) can be calculated from Equation 4 below.

Equation 4 is an expression for the zero crossing count, where X is the measured EMG signal, i is the order of the acquired signals, and N is the number of the acquired signals.

As such, the biometric state recognizer 400 may generate the clustered feature map previously stored in the storage 500 and the generated absolute based on the absolute difference standard deviation calculated from at least one of Equations 1 to 4. The differential standard deviation may be compared to determine the pattern of muscle movement or degree of muscle fatigue of the user.

Specifically, the biometric state recognition unit 400 generates an absolute difference standard deviation from at least one of Equations 1 to 4 for each channel. When the absolute difference standard deviation is generated from at least one of Equations 1 to 3, the biometric state recognition unit 400 may display the generated absolute difference standard deviation on the feature map extracted from the storage unit 500. have. Here, the feature map is a map in which a plurality of user movement patterns or muscle fatigue degrees are displayed as regions using an absolute difference standard deviation, and the biometric state recognition unit 400 generates an absolute difference standard deviation within a feature map. It is possible to determine the movement pattern or the degree of muscle fatigue of the user by determining whether it belongs to the area.

The dimension of the feature map is equal to the number of channels extracted by the signal extractor 200. For example, if the signal extractor 200 extracts the EMG signal for the p channels, the feature map will be formed in the p-dimensional. This will be described later in more detail with reference to FIGS. 4A to 4D.

The output unit 600 may receive the pattern recognized from the biometric state recognition unit 400 and transmit the pattern to the outside. In actual implementation, the output unit may be implemented as a USB port, Bluetooth, wired / wireless LAN, or the like.

3A to 3D are block diagrams illustrating a process of extracting a part of a plurality of EMG signals by the signal extractor according to embodiments of the present disclosure.

3A is a block diagram illustrating a process of extracting a part of a plurality of EMG signals by a signal extractor according to an exemplary embodiment.

Referring to FIG. 3A, the sensor unit 100 may transmit a plurality of EMG signals to the signal extractor 200. As time passes, the sensor unit 100 may repeatedly transmit the EMG signal to the storage unit 500. Therefore, the storage unit 500 may store time series data of the EMG signal transmitted from the sensor unit 100. Can be.

The signal extractor 200 may extract time series data on the EMG signal from the storage 500. The signal extractor 200 may analyze the extracted time series data to determine a channel having no change in the EMG signal over time. Thereafter, the signal extractor 200 extracts the EMG signals of the remaining channels except for the channel having no change in the EMG signal from the plurality of EMG signals received from the sensor unit 100 according to the above-described time. Can be sent to. As time passes, the biosignal recognition apparatus may stop collecting the EMG signal by determining that the muscle maintains a constant tension with respect to a channel without an EMG signal. Accordingly, the signal extractor 200 may extract some EMG signals from the plurality of EMG signals collected by the sensor unit 100.

3B is a block diagram illustrating a process of extracting a signal of a portion of a plurality of EMG signals by a signal extractor according to another embodiment of the present invention.

Referring to FIG. 3B, the sensor unit 100 may transmit a plurality of EMG signals to the signal extractor 200.

The storage unit 300 may store the EMG signal threshold value for determining the validity of the EMG signal.

The signal extractor 200 may compare the pre-stored EMG signal threshold value with the EMG signal strength received from the signal extractor 200 to extract only the EMG signal of a channel greater than the given threshold value and transmit the EMG signal to the ADC 300. .

When a user tries to perform a specific movement, muscles that do not contract or relax due to a specific movement may also have electromyography due to other external influences, and noise in determining the specific movement that the user intends. Can affect. Therefore, the EMG signal when the EMG signal is smaller than a specific threshold is processed as noise. Accordingly, the signal extractor 200 may extract only some EMG signals from the plurality of EMG signals collected by the sensor unit 100.

3C is a block diagram illustrating a process of extracting some EMG signals from a plurality of EMG signals by the signal extractor 200 according to another embodiment of the present invention.

Referring to FIG. 3C, the sensor unit 100 may collect a plurality of EMG signals and transmit them to the signal extractor 200. The storage unit 500 may store information on signal strength and signal extraction condition of the analog signal for each channel received from the sensor unit 100.

The signal extractor 200 may receive a plurality of EMG signals from the sensor unit 100. Thereafter, as described above with reference to FIG. 2, the biometric state recognition unit 400 may actually recognize the pattern about the movement intended by the user and store the pattern in the storage unit 500.

The storage unit 500 may store the information on the strength of the EMG signal in correspondence with the movement pattern or muscle fatigue of the user recognized by the biometric state recognition unit 400. As the recognition operation is repeated several times, a plurality of various movement patterns may be stored in the storage unit 500 in correspondence with muscle fatigue information and information on the strength of an EMG signal.

The signal extractor 200 may extract the movement pattern or the muscle fatigue information recognized from the storage 500 and information on the intensity of the EMG signal corresponding thereto. Information on the strength of the EMG signal extracted by the signal extractor 200 may be compared with the strength of the EMG signal newly received from the sensor unit 100.

The signal extractor 200 recognizes that the corresponding movement pattern or muscle fatigue information is input when the difference between the strength of the stored EMG signal and the intensity of the EMG signal newly received from the sensor 100 is less than or equal to a threshold having a predetermined value. EMG information of the channel corresponding to the movement pattern or the muscle depth information may be calculated. Accordingly, the signal extractor 200 may extract some EMG signals of the plurality of EMG signals, and the storage unit 500 may store signals that satisfy signal extraction conditions such as thresholds.

According to an additional aspect of the present invention, the signal extractor 200 may receive a plurality of signals received from the sensor unit 100 using a principal component analysis (PCA) method or an independent component analysis (ICA) method. A valid EMG signal may be extracted from the EMG signals. Principal component analysis method is a method for screening only EMG signals that are effective in a large number of data, and a group of data represented by P-type variables that are related to each other is converted into P-type characteristic values that are not related to each other by transformation of coordinate axes. It is a method of extracting only the most reverberant information of the original variable among the main components. In this way, it is possible to identify structures that exist within variables and to remove variables of low importance, that is, not belonging to factors.

3D is a block diagram illustrating a process of extracting some EMG signals from a plurality of EMG signals by the signal extractor 200 according to another exemplary embodiment.

Referring to FIG. 3D, the plurality of sensor units 100-1, 100-2, and 100-n may include a signal extractor 200, a switching unit 150, a user interface unit 180, a power supply unit 130, and an ADC unit. 300, the biometric state recognition unit 400 may be included.

The user interface unit 180 may receive an external signal for selecting a sensor to be activated from an external user. The user interface 190 may be implemented as a computer screen screen as shown in FIG. 3E and may be implemented as a touch screen or other electronic device.

The switching unit 150 may receive power from the power supply unit 130 and supply the power to the sensor units 100-1, 100-2, and 100-n. In addition, the switching unit 150 may receive an external signal for selecting a sensor to be activated from the user interface unit 180 and activate a sensor corresponding thereto. More specifically, the switching unit 150 connects only the wires corresponding to the selected sensor among the wires connected to the sensors 100-1, 100-2, 100-n from the power supply unit 130 based on an external signal, and the remaining wires. Can remain open.

3E is an exemplary view of implementing a user interface unit according to an embodiment of the present invention.

As shown in FIG. 3E, the user may select a portion of the body to measure muscle movement. In FIG. 3D, the user measures his left arm as an example.

When the user inputs an external signal to the user interface unit 180 to measure the movement of his left arm, the user interface unit 180 transmits it to the switching unit 150 and the switching unit 150 supplies a sensor corresponding to the left arm. Enable and disable the remaining sensors. Accordingly, the signal extractor 200 may extract some EMG signals of a plurality of EMG signals.

According to the embodiments related to FIGS. 3A to 3E, the signal extractor 200 may extract some EMG signals of a plurality of EMG signals, thereby dramatically reducing the time required to process data and a memory required for data processing. Can be reduced. In addition, the effect of the error can be reduced by excluding the unextracted EMG signal, so that more accurate movement pattern and muscle fatigue recognition work can be performed.

 4A to 4D are graphs for explaining a process of forming and clustering a feature map that is pre-stored in the storage unit 500 and used to recognize a user's movement pattern or a muscle fatigue of the user. 4A to 4D, description will be made based on an embodiment in which the signal extractor 200 extracts an EMG signal for two channels.

The signal extractor 200 may transmit the EMG signals collected in the channel 1 and the channel 2 of the sensor unit 100 to the ADC unit 300 to perform an analog-to-digital conversion operation on the analog EMG signal.

The biometric state recognition unit 400 may store the digital value for each channel over time in the storage 500. The biometric state recognition unit 400 may extract a time-series digital value stored in the storage unit 500 for a predetermined time when the sensor unit 100 does not collect any EMG signals due to reaching a predetermined time. .

As described above, the biometric state recognizer 400 may generate an absolute difference standard deviation, an absolute integral value, an accumulated integral value, and a zero crossing frequency with respect to a time series digital value. The feature map may be generated using the generated digital value. Can be generated.

4A is an example of a feature map. Since the signal extracting unit 200 extracts signals with respect to two channels, it is configured in two dimensions. For convenience, the x and y axes will be referred to as channel 1 and channel 2, respectively. In the present embodiment, the EMG signal is applied to the EMG signal over a predetermined time for a predetermined time in the case where the user's wrist is left, right, up, down, and rest. Collected. When at least one of Equations 1 to 4 is applied to the time series data of the EMG signal collected 30 times for each channel, 30 absolute difference standard deviations are generated. In addition, since the experiment was performed on two channels, a total of 30 two-dimensional ordered pairs of absolute difference standard deviations are generated. 30 ordered pairs obtained through the above-described process are displayed on the feature map.

For example, if the subject faces the wrist upward, an absolute differential standard deviation of 0.8 will be generated in channel 1, and if an absolute differential standard deviation of 0.2 is generated in channel 2, it will be at the point where the x-axis 0.2 and the y-axis 0.8 intersect. To display.

Referring to FIG. 4A, thirty points are respectively shown for the operation of the test subject, which is distinguished by a circle in a stationary state, a square in the top, a diamond in the bottom, a cross in the right, and a point in the left according to the shape of the point. . Since 30 experiments were performed on each of the five wrist movements, a total of 150 points are shown, thereby forming on the feature map.

Once the feature map is formed, clustering is performed on the feature map. Clustering can be performed by k-near method, quadratic discriminant analysis (QDA), linear discrimineant analysis (LDA) algorithm, and FIG. 4B is k-near neighbor method, FIG. 4C is QDA algorithm, and FIG. 4D is LDA algorithm. Graph showing clustered results. It can be seen that the area of the wrist movement is shown based on the data of the points shown in the feature map.

The biometric state recognition unit 400 extracts the clustered feature map obtained through the above-described process from the storage unit 500, and matches the absolute difference standard deviation generated by the clustered feature map to determine the corresponding region, and determines the movement of the wrist. The pattern may be recognized, and the degree of fatigue of the user's wrist use may be determined by referring to the pattern information of the recognized wrist movement.

In FIGS. 4B to 4D, clustering is performed only by the k-near method, the QDA, and the LDA algorithm, but this is merely an example. Clustering may be performed by various clustering methods such as the K-means and Fuzzy C-means methods. have.

5 is a block diagram illustrating an apparatus for recognizing a biosignal according to still another example of the present invention.

As shown in FIG. 5, the biosignal recognition apparatus includes a sensor unit 100, a signal extractor 200, an ADC unit 300, a biometric state recognition unit 400, a storage unit 500, and an output unit 600. ) May be included. However, description of components having the same functions as those described in FIG. 2 among the components of the present exemplary embodiment will be omitted.

Unlike the embodiment described with reference to FIG. 2, the ADC operation is performed before the operation of extracting some EMG signals from the EMG signals extracted from the sensor unit 100 is performed. As shown in FIG. 5, six signals extracted from the sensor unit 100 are transmitted to the ADC unit 300, and all of the EMG signals corresponding to the six channels are converted into digital signals to generate digital values.

The signal extractor 200 may extract digital values of some of the six digital values. The method of extracting some digital values from the signal extractor 200 is as described with reference to FIGS. 3A to 3D.

Fast Fourier transformation (FFT) may be performed on the digital EMG signal extracted by the ADC unit 300 according to an embodiment of the present invention. The digital value generated by performing the FFT includes frequency information about the EMG signal.

The signal extractor 200 may extract some digital values from the plurality of digital values based on the frequency information of the digital values. The signal extractor 200 may be achieved through Equations 5 and 6 below. Equations 5 and 6 are equations for representing a center frequency and an average frequency used when extracting some digital frequency values from a plurality of digital frequency values.

The signal extractor 200 may extract digital values of some of the six digital values. The method of extracting some digital values from the signal extractor 200 is as described with reference to FIGS. 3A to 3D. Fast Fourier transformation (FFT) may be performed on the digital EMG signal extracted by the ADC unit 300 according to an embodiment of the present invention. The digital value generated by performing the FFT includes frequency information about the EMG signal. The signal extractor 200 may extract some digital values from the plurality of digital values based on the frequency information of the digital values. The signal extractor 200 may be achieved through Equations 5 and 6 below. Equations 5 and 6 are equations for representing a center frequency and an average frequency used when extracting some digital frequency values from a plurality of digital frequency values.

Equation 5 is an equation for the center frequency, where s (f) is an area density function corresponding to the frequency f of the digital value. And fmed means central frequency. Therefore, the formula for the center frequency can be defined as the center frequency, which is 1/2 of the total area of the lower part of the area density function.

The equation for the average frequency can be achieved from Equation 6 below.

Equation 5 is an expression for the average frequency, where fi denotes the i-th frequency, and Ii denotes the intensity corresponding to fi. In addition, fmean means the average frequency. Therefore, the formula for the average frequency may be referred to as the average value measured based on the intensity as the frequency and the frequency as the variable.

The storage unit 500 may store a predetermined threshold range for the center frequency and the average frequency. For example, in the case where the user performs an exercise of contracting the biceps for 5 seconds and relaxing for 5 seconds, the median frequency generated using Equations 4 and 5 is 72.98 ± 4.15 and the average frequency is 90.14 ± 4.17. The signal extractor 200 may extract only a digital value corresponding to a case in which the signal extractor 200 exists within a preset threshold range using existing measured data.

Until now, the respective components of the biosignal recognition apparatus for selectively extracting some of the EMG signals detected from the plurality of sensors to determine the degree of movement of the user or the degree of muscle fatigue of the user have been described in detail. Hereinafter, a method of determining a user's movement pattern or muscle fatigue degree by extracting an EMG signal of some of the EEG signals detected by the plurality of sensors in the biometric device according to the present invention will be described in detail.

6 is a flowchart illustrating a method of recognizing a biosignal according to an exemplary embodiment.

As shown in FIG. 6, the biosignal recognition apparatus collects a plurality of EMG signals sensed from a plurality of sensors attached to a user's living body (S510). Thereafter, the biosignal recognition apparatus selectively extracts only a portion of signals collected from the plurality of collected EMG signals before the ADC operation is performed on the collected plurality of EMG signals (S520). According to an embodiment, the biosignal apparatus may compare the EMG signal thresholds for determining the validity of the EMG signal previously stored in the storage unit with a plurality of EMG signals sensed by the plurality of sensors, as shown in FIG. 3A. Only EMG signals related to the movement pattern above the signal threshold may be extracted.

According to another embodiment of the present disclosure, the biosignal device may include a pattern previously stored in the storage unit of the EMG signal sensed by the sensors based on a correspondence between the EMG signal previously stored in the storage unit and the movement pattern of the living body, as shown in FIG. 3B. Only EMG signals corresponding to the E can be extracted. In detail, a plurality of various patterns may be stored in the storage unit in correspondence with information on the strength of the EMG signal. Therefore, when the EMG signals are sensed from the plurality of sensors and the plurality of EMG signals are collected, the biosignal device may extract only EMG signals corresponding to patterns previously stored in the storage unit from the collected EMG signals.

According to another embodiment, the biosignal recognition apparatus may use an effective EMG signal among EMG signals sensed from a plurality of sensors using a Principal Component Analysis (PCA) method or an Independent Component Analsis (ICA) method. Only bays can be extracted. Since the description of the principal component analysis method has been described above in detail, a detailed description thereof will be omitted. According to various embodiments of the present disclosure, when some EMG signals of the EMG signals sensed by the plurality of sensors are extracted, the biosignal recognition apparatus converts the extracted EMG signals into digital values through an ADC operation (S530). Thereafter, the biosignal recognition apparatus processes the digital value converted for each channel through the ADC operation to be stored in the storage unit as time series data. When the digital signal converted for each channel is stored in the storage unit over a predetermined time, the biosignal recognition apparatus calculates an absolute difference standard deviation with respect to the time-series digital values previously stored in the storage unit. Such a biosignal recognition apparatus may calculate an absolute difference standard deviation with respect to time series data values previously stored in a storage unit by using at least one of Equations 1 to 4 described above. Here, the absolute difference standard deviation is calculated using at least one of the absolute integral value, the cumulative integral value, and the zero crossing count, and the absolute integral average value, the cumulative integral value, and the zero crossing count counting the absolute difference standard deviation are As described above, this may be achieved through Equations 1 to 4.

When the digital value is calculated through one of Equations 1 to 4, the biosignal recognition apparatus compares the absolute difference standard deviation generated by the number of channels with the clustered feature map previously stored in the storage unit, and compares the result. In operation S540, the pattern of the user's living body movement may be recognized or muscle fatigue information may be determined. Thereafter, the biosignal recognition apparatus stores the recognized movement pattern information or the information on the degree of muscle fatigue degree in a storage unit or outputs the external signal to be transmitted to another application (S550).

7 is a flowchart for explaining another biosignal recognition method of the present invention.

As illustrated in FIG. 7, the biosignal recognition apparatus collects a plurality of EMG signals sensed from sensors attached to various parts of the user's body (S710). Here, the plurality of EMG signals sensed from the plurality of sensors are analog signals. Therefore, when the plurality of analog EMG signals sensed by the plurality of sensors are collected, the biosignal recognition apparatus converts the plurality of analog EMG signals into digital values through an ADC operation (S720). Thereafter, the biosignal recognition apparatus extracts a digital EMG signal corresponding to a digital value of some of the digital values converted from the analog EMG signal (S730). Since a method of extracting only some EMG signals has been described above in detail, a detailed description of EMG signal extraction methods will be omitted.

According to a predetermined condition, when some digital EMG signals are extracted, the biosignal recognition apparatus generates a digital frequency value by performing an FFT on the extracted digital EMG signals. Thereafter, the biosignal recognition apparatus may extract some digital frequency values from the plurality of digital frequency values through Equations 5 and 6 as described above. When some digital frequency values are extracted from the plurality of digital frequency values through Equation 5 and Equation 6, the biosignal recognition apparatus recognizes a movement pattern of the living body based on the extracted some digital frequency values or according to the movement pattern. The degree of muscle fatigue can be determined (S740). As such, when the user recognizes the movement pattern or the degree of muscle fatigue according to the movement pattern is determined in step S740, the biosignal recognition apparatus stores the recognized movement pattern information or the information on the degree of muscle fatigue in the storage unit or other information. Outputs to the outside to be transmitted to the application (S750).

In the above described and described with respect to the preferred embodiment of the present invention anyone of ordinary skill in the art belongs to the preferred embodiment of the present invention within the scope without departing from the spirit and scope of the present invention Of course, it can be changed in various ways. Accordingly, it is intended that the present invention cover the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.

100: sensor unit 200: signal extraction unit
300: ADC unit 400: biological state recognition unit
500: storage unit 600: output unit

Claims (24)

A plurality of sensor units respectively attached to a living body to sense an EMG signal;
A signal extractor configured to selectively extract some of a plurality of EMG signals sensed from the plurality of sensor units, respectively; And
In the bio-signal recognition apparatus comprising: a biometric state recognition unit for recognizing at least one of the movement pattern or muscle fatigue of the living body based on the extracted EMG signal,
And a storage unit configured to store an EMG signal threshold value for determining the validity of the EMG signal.
The signal extractor,
And extracting only an EMG signal which is greater than or equal to the EMG signal by comparing the EMG signal threshold value extracted from the storage unit with the EMG signal. delete The method of claim 1,
An analog-to-digital converter (ADC) unit for performing an analog-to-digital conversion on the EMG signal extracted from the signal extractor to generate a digital value for the EMG signal;
Biological signal recognition device further comprising. The method of claim 3, wherein
The storage unit may further store a feature map in which an area corresponding to at least one of the movement pattern or the muscle fatigue level is displayed as a digital value.
And the biological state recognition unit recognizes at least one of the movement pattern and the muscle fatigue degree based on the digital value and the feature map extracted from the storage unit. 5. The method of claim 4,
The storage unit further stores time series data corresponding to the digital value for a preset time,
The digital value that is used by the biometric state recognition unit to recognize at least one of the movement pattern or the muscle fatigue is calculated using at least one of absolute difference standard deviation, absolute integral average value, cumulative integral value, and zero crossing count. Biometric signal recognition device. The method of claim 1,
The storage unit further stores a correspondence relationship between the EMG signal and the movement pattern or the muscle fatigue,
The signal extractor,
And extracting at least one EMG signal based on a corresponding relationship between the EMG signal received from the sensor unit and the EMG signal extracted from the storage unit and the movement pattern or muscle fatigue. The method of claim 1,
The signal extractor,
A biosignal recognition apparatus, comprising extracting a valid EMG signal from the plurality of EMG signals using principal component analysis (PCA) or independent component analysis (ICA). 5. The method of claim 4,
The biometric state recognition unit clusters using at least one of a k-nearest neighbor method, a fuzzy c-means method, a quadratic discriminant analysis (QDA), and a linear discriminant analysis (LDA) algorithm. The apparatus for recognizing a biosignal, characterized in that for recognizing a pattern using the feature map. 5. The method of claim 4,
And an output unit configured to output the movement pattern or the muscle fatigue information received from the biometric state recognition unit to the outside using a wired or wireless communication device. The method of claim 1,
A user interface unit configured to receive an external signal for selecting some of the plurality of sensor units;
A power supply unit supplying power to the biosignal recognition apparatus; And
And a switching unit receiving power from the power supply unit based on the external signal to supply power to the corresponding some sensor units.
The signal extracting unit extracts an EMG signal from the sensor unit supplied with power from the switching unit. The method of claim 1,
And an ADC unit receiving the EMG signal from the plurality of sensor units, converting an analog value into a digital value and transmitting the analog value to a digital signal.
And the signal extracting unit extracts a part of digital values from the plurality of digital values by using an absolute difference standard deviation calculated from a center frequency or an average frequency related equation. The method of claim 11,
The storage unit further stores a threshold range for determining the validity of the digital value,
A digital frequency value is generated by performing FFT (fast fourier transformation) on the digital EMG signal output from the ADC,
The signal extractor,
And extracting the digital frequency value included in the threshold range extracted from the storage unit. Sensing a plurality of EMG signals to a living body for a predetermined time using the plurality of sensors;
Selectively extracting some of the sensed EMG signals; And
Recognizing at least one of the movement pattern or muscle fatigue of the living body based on the extracted EMG signal, comprising:
Selectively extracting some of the plurality of EMG signals,
And comparing only the EMG signal threshold value for determining the validity of the pre-stored EMG signal with the EMG signal, and extracting only an EMG signal related to a movement pattern equal to or greater than the EMG signal threshold value. delete The method of claim 13,
And performing a digital-to-digital conversion on the extracted EMG signal to generate a digital value for the EMG signal. The method of claim 15,
And the digital value is calculated using at least one of absolute difference standard deviation, absolute integral average value, cumulative integral value, and zero crossing count. The method of claim 15,
Recognizing at least one of the movement pattern or the muscle fatigue based on the extracted partial EMG signal,
And recognizing the movement pattern or the muscle fatigue on the basis of the digital value and a feature map in which regions corresponding to the previously stored movement pattern or muscle fatigue are displayed as digital values. The method of claim 17,
The feature map is clustered using at least one of a k-nearest neighbor method, a fuzzy c-means method, a quadratic discriminant analysis (QDA), and a linear discriminanat analysis (LDA) algorithm. A biosignal recognition method, characterized in that. The method of claim 13,
Selectively extracting a part of the sensed plurality of EMG signals,
And extracting only an EMG signal corresponding to the stored pattern from the collected EMG signals based on a corresponding relationship between the pre-stored EMG signal and the movement pattern of the living body. The method of claim 19,
Outputting the pattern using a wired or wireless communication device;
Biological signal recognition method further comprises. The method of claim 13,
Selectively extracting a part of the sensed plurality of EMG signals,
A biosignal recognition method comprising extracting a valid EMG signal from the plurality of EMG signals using principal component analysis (PCA) or Independent Component Analysis (ICA). The method of claim 13,
Receiving an external signal for selecting some of the plurality of sensors; And
Based on the external signal, supplying power to the corresponding some sensor;
Selectively extracting a part of the sensed plurality of EMG signals,
Extracting an EMG signal from the powered sensor. Sensing a plurality of EMG signals to a living body for a predetermined time using the plurality of sensors;
Generating a plurality of digital frequency values for the plurality of signals;
Selectively extracting some of the plurality of digital frequency values; And
Recognizing at least one of the movement pattern or muscle fatigue of the living body based on the extracted digital frequency value, comprising:
The digital frequency value is generated by FFT,
Selectively extracting a part of the plurality of digital frequency values,
And selectively extracting a digital frequency value included in a threshold range for determining the validity of the previously stored digital frequency value. delete

Images