WO2020186651A1 - 基于脑电意念的智能运动耳机及其实现方法与系统 - Google Patents

基于脑电意念的智能运动耳机及其实现方法与系统 Download PDF

Info

Publication number
WO2020186651A1
WO2020186651A1 PCT/CN2019/092880 CN2019092880W WO2020186651A1 WO 2020186651 A1 WO2020186651 A1 WO 2020186651A1 CN 2019092880 W CN2019092880 W CN 2019092880W WO 2020186651 A1 WO2020186651 A1 WO 2020186651A1
Authority
WO
WIPO (PCT)
Prior art keywords
module
eeg
headset
dry electrode
music
Prior art date
Application number
PCT/CN2019/092880
Other languages
English (en)
French (fr)
Inventor
杜安明
黄海平
肖甫
程琨
王汝传
沙超
李家东
李欣祥
Original Assignee
南京邮电大学
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 南京邮电大学 filed Critical 南京邮电大学
Publication of WO2020186651A1 publication Critical patent/WO2020186651A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R1/00—Details of transducers, loudspeakers or microphones
    • H04R1/10—Earpieces; Attachments therefor ; Earphones; Monophonic headphones
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • H04R3/12—Circuits for transducers, loudspeakers or microphones for distributing signals to two or more loudspeakers
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Definitions

  • the invention relates to an intelligent sports earphone based on EEG idea and its realization method and system. It is based on EEG signal acquisition, EEG data deep learning model design and embedded software and hardware design realization scheme, aiming to realize EEG based idea Smart sports headphones with deep learning belong to the intersection of brainwave technology and embedded systems.
  • the purpose of the present invention is to provide a smart sports earphone based on brain electricity idea and its realization method and system, realize the control of the earphone, and solve the problem that manual control is required for listening to music during exercise in the prior art.
  • a smart sports headset based on brain electricity ideas including a power supply and a headset body.
  • the headset body includes a headset, a headset control module and a sound unit. Both ends of the headset are equipped with sound units.
  • the headset control module is built into the headset. It also includes an EEG signal acquisition module, an EEG signal preprocessing module, an identification module and a storage module.
  • the EEG signal acquisition module, EEG signal preprocessing module, identification module and storage module are respectively built into the headset, and the EEG signal acquisition
  • the module includes dry electrode sensor 1a, dry electrode sensor 1b, dry electrode sensor 1c and dry electrode sensor 1d, dry electrode sensor 1a, dry electrode sensor 1b, dry electrode sensor 1c and dry electrode sensor 1d, respectively Set on the headwear, the EEG signal acquisition module is connected to the recognition module through the EEG signal preprocessing module, and the recognition module is connected to the storage module.
  • the eSIM module is built in the headset, the eSIM module is connected to the headset control module, the eSIM module uses the eSIM NB-IOT communication module M5310, and the identification module uses the STM32F072RB chip.
  • the dry electrode sensor 1a and the dry electrode sensor 1d, the dry electrode sensor 1b and the dry electrode sensor 1c are bipolar in parallel, respectively, and the dry electrode sensor 1a and the dry electrode sensor 1d are respectively arranged at both ends of the head wear ,
  • the dry electrode sensor 1b and the dry electrode sensor 1c are respectively arranged on both ends of the headwear and are arranged between the dry electrode sensor 1a and the dry electrode sensor 1d; the dry electrode sensor 1a and the dry electrode sensor 1d, dry Both the state electrode sensor 1b and the dry state electrode sensor 1c adopt active dry state electrode sensors.
  • the identification module uses the STM32F072RB chip.
  • a method for implementing smart sports earphones based on brain electricity ideas described in any one of the above includes the following steps:
  • EEG signal acquisition using the dry electrode EEG acquisition technology
  • the dry electrode sensors 1a, 1b, 1c, 1d collected by the four single data channels in the EEG signal acquisition module are used to collect the brain through the set sampling frequency.
  • the EEG signal preprocessing module receives the EEG signals collected by the EEG acquisition module, and the EEG signal preprocessing module filters and amplifies the EEG signals;
  • step S4 Use the backpropagation algorithm to bring the classification results obtained in step S3 into the feedforward neural network training, obtain the number of hidden layer neurons n and the weight matrix W, and obtain the EEG based on the feedforward neural network used in this method Feature deep learning model;
  • step S5 Recognize the EEG signals collected in real time, and directly preset the parameters obtained in step S4 in the recognition module, that is, preset the EEG feature deep learning model established in S4 in the recognition module, and record the movement process of the athlete
  • the EEG signal data collected in real-time in the computer is preprocessed by the preprocessing module and input to the recognition module to obtain the corresponding classification results;
  • step S7 performing call control through the eSIM module to realize the control of receiving incoming calls, answering calls, and hanging up calls, and sending relevant control information to the headset control module.
  • step S3 an EEG feature extraction method combining independent component analysis ICA and wavelet analysis WY is used to perform feature extraction to obtain classification results, specifically,
  • step S31 Perform dyadic wavelet transformation on the EEG signal preprocessed in step S2, and the definition formula of the dyadic wavelet transformation of the brain wave signal is: Among them, c i, j dyadic wavelet coefficients, k is an integer, Is the wavelet function, Is a dyadic wavelet function, and its inverse transformation formula is: Among them, x j (t) represents the component of the signal x(t) at a certain scale (2 j );
  • f s is the sampling frequency
  • n and m are the selected subband range
  • EEG signals There are 5 main frequency bands for EEG signals: Delta (0-4Hz), Theta (4-8Hz), Alpha (8-15Hz), Beta (15-30Hz) and Gamma (30-60Hz).
  • step S4 the back propagation algorithm is used to train the feedforward neural network learning model, specifically:
  • z l is the lth input
  • W is the weight matrix from the lth layer to the l+1th layer
  • b is the bias from the lth layer to the l+1th layer
  • x is the value of each neuron in the lth layer
  • y is The value of each neuron in the l+1 layer
  • J is the Jacobian matrix
  • ⁇ is the product of the Hamiltonian, a n represents the n-th layer;
  • the modified weight matrix gradient is:
  • ⁇ W′ n is the gradient of the weight matrix of the nth layer after correction
  • ⁇ W n is the gradient of the weight matrix of the nth layer before correction
  • the corrected bias gradient is:
  • ⁇ b' is the bias gradient of the nth layer after correction
  • ⁇ b n is the bias gradient of the nth layer before correction
  • the number of hidden layer neurons is selected according to the empirical formula:
  • n is the number of hidden layer neurons
  • n i is the number of input neurons
  • n 0 is the number of output neurons
  • a is a constant between 1-10.
  • step S47 Add labels to the independent components extracted in step S3, and input them into the back-propagation feedforward neural network model according to the above steps, and then the parameters in the network model can be obtained by training.
  • the earphone control module receives the embedded instructions from the recognition module, and sends instructions to the sound unit according to the following priority order, specifically,
  • the headset controls The module enters the waiting state until it receives the embedded instruction to answer the call or the embedded instruction to hang up the call from the recognition module;
  • the headset control module sends an instruction to answer the call to the eSIM module, and sends the information to play the call voice to the sound unit, and then continues to enter the waiting state; the eSIM module chooses to answer the call, and The controller makes the headset start to transmit the call voice;
  • the headset control module sends an instruction to hang up the phone to the eSIM module and an instruction to continue the previous task to the sounding unit;
  • an embedded instruction to enter the smart music mode is received in the non-smart music mode, an instruction to play music from the smart music control module is sent to the sound unit; if an embedded instruction to exit the smart music mode is received in the smart music mode , Then send an instruction to play music from the storage module to the sound unit;
  • An implementation system of an EEG-based smart sports headset using any one of the above-mentioned implementation methods of an EEG-based smart sports headset including an EEG signal acquisition module, an EEG signal preprocessing module, and an identification module , ESIM module and storage module,
  • EEG signal acquisition module for EEG signal acquisition, using dry electrode EEG acquisition technology, the dry electrode sensors 1a, 1b, 1c, 1d collected by the four single data channels in the EEG signal acquisition module are set by Sampling frequency for EEG signal collection;
  • EEG signal preprocessing module preprocess the collected EEG signals, the EEG signal preprocessing module receives the EEG signals collected by the EEG acquisition module, and the EEG signal preprocessing module filters and amplifies the EEG signals ;
  • Recognition module Obtain the classification results through the EEG feature extraction model, and use the EEG feature extraction method combining independent component analysis ICA and wavelet analysis WY to extract the features of the EEG signal preprocessed in step S2 to obtain the classification results; establish the brain Electrical feature deep learning model: use the back propagation algorithm to train the feedforward neural network learning model; recognize the EEG signals collected in real time, and directly preset the parameters obtained in steps S3 and S4 in the recognition module to determine the movement process of the athlete
  • the EEG signal data collected in real time in the computer is preprocessed by the preprocessing module, and then input to the recognition module to obtain the corresponding classification result; send instructions to the headset control module, and the classification result obtained after the processor of the recognition module is calculated, and The classification results in the storage module are compared, the corresponding embedded instructions are obtained, and the embedded instructions are sent to the headset control module;
  • eSIM module call control through the eSIM module to realize the control of receiving incoming calls, answering calls, and hanging up calls, and send relevant control information to the headset control module;
  • Storage module used to store the classification results of EEG signals and the corresponding embedded instructions.
  • an intelligent music control module receiving real-time heart rate data collected by the heart rate sensor, so that the volume and frequency of the music will change accordingly with the changes in the athlete's real-time heart rate data, specifically,
  • the exercise intensity represented by the current heart rate is analyzed in real time.
  • the relationship between the exercise intensity of the athlete and the real-time heart rate of the athlete is: Where R t is the real-time heart rate of aerobic exercise, R m is the maximum heart rate of aerobic exercise, R q is the resting heart rate, Y is the age of the exerciser, and Q R is the real-time exercise intensity of the exerciser measured by the heart rate. ;
  • brain waves to measure the growth rate of the relative power of the brain waves to measure the exercise intensity of the exerciser, which is defined as follows:
  • P is the relative power of the brain wave when the athlete is at rest
  • ⁇ P is the change in the relative power of the brain wave when the athlete is exercising
  • f 0 represents the initial music frequency
  • v 0 represents the initial music volume
  • This kind of smart sports earphone based on EEG idea can realize the control of the earphone body through the collection and preprocessing of EEG signals, and the structure design is reasonable and easy to use. Realize the acquisition of EEG signals through the EEG signal acquisition module, and realize the preprocessing of the collected EEG signals through the EEG signal preprocessing module, realize the recognition of EEG signals through the recognition module, and then send control instructions to the headset control The module realizes the control of the sound unit.
  • This smart sports headset based on brain electricity ideas creatively embeds the eSIM module into the wireless headset, and through its own design of the headset control module, it is possible to use brain electricity ideas to control and answer important calls without carrying a mobile phone, so that The user completely gets rid of the fetters of the mobile phone when exercising.
  • This smart sports headset based on EEG idea and its implementation method and system, pre-input feature extraction and neural network parameters into the processing module, eliminating the preprocessing and feature extraction steps in traditional EEG processing , Can quickly and accurately obtain EEG characteristics and classify.
  • the present invention adopts a feature extraction method combining wavelet analysis and independent component analysis, and adopts a back-propagation feedforward neural network to make the classification result more accurate.
  • the present invention realizes the integrated service of EEG signal collection, analysis and processing, classification result conversion, wireless earphone control, incoming call answering, and smart music, and finally realizes the smart sports earphone based on EEG idea.
  • This kind of smart sports earphone based on brain electricity idea and its realization method and system, creatively combine the control of brain wave technology and wireless plug-in earphone technology to realize a brand new earphone control method, which can be well Help athletes get rid of the trouble of disturbing the rhythm of exercise due to manual control of earphones.
  • Fig. 1 is a schematic diagram of the structure of the earphone body in the embodiment.
  • FIG. 2 is an explanatory block diagram of an implementation system of a smart sports earphone based on EEG idea according to an embodiment of the present invention
  • Fig. 3 is a schematic diagram of the process of feature extraction and classification and identification parameter determination in the embodiment.
  • Figure 4 is an ASM diagram of the headset control module in the embodiment.
  • a smart sports headset based on brain electricity ideas includes a power supply and a headset body.
  • the headset body includes a headset, a headset control module and a sound unit. Both ends of the headset are equipped with sound units.
  • the headset control module is built in
  • the headset also includes an EEG signal acquisition module, an EEG signal preprocessing module, an identification module, and a storage module.
  • the EEG signal acquisition module, EEG signal preprocessing module, identification module and storage module are respectively built into the headset.
  • the EEG signal acquisition module includes dry electrode sensor 1a, dry electrode sensor 1b, dry electrode sensor 1c and dry electrode sensor 1d, dry electrode sensor 1a, dry electrode sensor 1b, dry electrode sensor 1c and dry electrode sensor 1c.
  • the electrode sensors 1d are respectively arranged on the headwear, the brain electrical signal acquisition module is connected to the identification module through the brain electrical signal preprocessing module, and the identification module is connected to the storage module.
  • This kind of smart sports headset based on EEG idea can realize the control of the headset body, realize the collection of EEG signals through the EEG signal acquisition module, and realize the preprocessing of the collected EEG signals through the EEG signal preprocessing module , Realize the recognition of EEG signals through the recognition module, and then send control instructions to the earphone control module to control the sound unit.
  • This smart sports headset based on brain electricity ideas creatively embeds the eSIM module into the wireless headset, and through its own design of the headset control module, it is possible to use brain electricity ideas to control and answer important calls without carrying a mobile phone, so that The user completely gets rid of the fetters of the mobile phone when exercising.
  • This smart sports headset based on brain electricity ideas also includes an eSIM module, which is built into the headset, the eSIM module is connected to the headset control module, the eSIM module uses the eSIM NB-IOT communication module M5310, and the identification module uses the STM32F072RB chip.
  • the embodiment preferably adopts the eSIM NB-IOT communication module M5310 released by China Mobile.
  • the module has the smallest size among similar products in the world, only 19 ⁇ 18.4 ⁇ 2.7mm, saving the board area by more than 30%.
  • the module uses HiSilicon Hi2110 chip, supports eSIM technology and OneNET platform protocol.
  • the recognition module uses the STM32F072RB chip.
  • the recognition module uses the STM32F072RB chip.
  • the STM32F072RB chip is a 32-bit microprocessor based on ARM Cortex-M7. It has a 64-bit AXI system bus interface, an instruction and data cache, 64-bit instruction tightly coupled memory (ITCM), dual 32-bit data tightly coupled memory (DTCM), and has extremely high computing performance.
  • the dry electrode sensor 1a and the dry electrode sensor 1d, the dry electrode sensor 1b and the dry electrode sensor 1c are bipolar parallel respectively, and the dry electrode sensor 1a and the dry electrode sensor 1d are respectively arranged in the head
  • the dry electrode sensor 1b and the dry electrode sensor 1c are respectively arranged on the two ends of the headset and between the dry electrode sensor 1a and the dry electrode sensor 1d; the dry electrode sensor 1a and the dry electrode sensor 1d
  • the sensor 1d, the dry electrode sensor 1b, and the dry electrode sensor 1c all use active dry electrode sensors.
  • the brain wave data acquisition uses the stem electrode brain electrical data acquisition technology.
  • dry electrode acquisition technology uses microneedle technology, ultra-high input impedance amplifier and other technologies to get rid of the dependence on conductive media and is simple and easy to use.
  • the EEG signal acquisition module of the embodiment is composed of four dry electrode sensors 1a, 1b, 1c, and 1d.
  • 1a and 1d adopt bipolar parallel connection and are located on both sides of the scalp area above the wireless card earphone.
  • 1b and 1c also use bipolar parallel connection, which are located below the preprocessing module and the storage module respectively.
  • the sampling frequency of the EEG signal acquisition module is 128 Hz.
  • the two EEG signals collected are the potential difference between the two groups of active electrodes.
  • the dry electrode sensor adopts the active dry electrode of g.SAHARA model produced by g.tec company.
  • a method for implementing smart sports earphones based on EEG ideas described in any one of the above, as shown in Figure 2, includes the following steps:
  • EEG signal acquisition using dry electrode EEG acquisition technology, dry electrode sensors 1a, 1b, 1c, 1d collected by the four single data channels in the EEG signal acquisition module through the set sampling frequency Perform EEG signal acquisition.
  • preprocessing preprocessing the collected EEG signals, the EEG signal preprocessing module receives the EEG signals collected by the EEG acquisition module, and the EEG signal preprocessing module filters and amplifies the EEG signals;
  • the 2-lead EEG signal collected in step S1 is sent to the EEG signal preprocessing module through the USB interface.
  • the EEG signal preprocessing module performs filtering and amplification of EEG signals.
  • the EEG signal preprocessing module includes the first stage amplifier, 50Hz notch filter, low pass filter, band pass filter, and second stage amplifier in turn.
  • Classification Obtain the classification result through the EEG feature extraction model, and use the EEG feature extraction method combining independent component analysis ICA and wavelet analysis WY to perform feature extraction on the EEG signal preprocessed in step S2 to obtain the classification result;
  • step S3 the EEG feature extraction method combining independent component analysis ICA and wavelet analysis WY is used to perform feature extraction to obtain classification results, as shown in Figure 3, specifically,
  • step S31 Perform dyadic wavelet transformation on the EEG signal preprocessed in step S2, and the definition formula of the dyadic wavelet transformation of the brain wave signal is:
  • the inverse transformation formula is: Among them, x j (t) represents the component of the signal x(t) at a certain scale (2 j );
  • f s is the sampling frequency
  • n and m are the selected subband range
  • EEG signals There are 5 main frequency bands for EEG signals: Delta (0-4Hz), Theta (4-8Hz), Alpha (8-15Hz), Beta (15-30Hz) and Gamma (30-60Hz).
  • the extended Infomax algorithm that is, the extended maximum entropy algorithm, is to solve the separation matrix W according to the natural gradient method.
  • the target criterion is that the target entropy is extremely large after the independent component passes through the nonlinear link.
  • the algorithm can be applied to both super-Gaussian and sub-Gaussian signals, and avoids the problem of matrix inversion, which improves the convergence speed.
  • the adjustment formula of the unmixing matrix of the extended maximum entropy algorithm is: Among them, I is the identity matrix, K is the diagonal matrix, and the diagonal elements are as follows:
  • ICA independent component analysis
  • PCA principal component analysis
  • WY wavelet analysis
  • the method in the embodiment adopts the back propagation algorithm to train the feedforward neural network learning model.
  • step S4 the back propagation algorithm is used to train the feedforward neural network learning model, specifically:
  • ⁇ is the product of the Hamiltonian, a n represents the n-th layer;
  • the modified weight matrix gradient is:
  • ⁇ W′ n is the gradient of the weight matrix of the nth layer after correction
  • ⁇ W n is the gradient of the weight matrix of the nth layer before correction
  • the corrected bias gradient is:
  • the number of hidden layer neurons is selected according to the empirical formula:
  • n is the number of hidden layer neurons
  • n i is the number of input neurons
  • n 0 is the number of output neurons
  • a is a constant between 1-10.
  • step S47 Add labels to the independent components extracted in step S3, and input them into the back-propagation feedforward neural network model according to the above steps, and then the parameters in the network model can be obtained by training.
  • Control send instructions to the headset control module, and compare the classification results obtained by the processor of the recognition module with the classification results in the storage module to obtain the corresponding embedded instructions, and send the embedded instructions to the headset control Module.
  • the embodiment further includes step S7, performing call control through the eSIM module to realize the control of receiving incoming calls, answering calls, and hanging up calls, and sending relevant control information to the headset control module.
  • step S7 performing call control through the eSIM module to realize the control of receiving incoming calls, answering calls, and hanging up calls, and sending relevant control information to the headset control module.
  • wireless card headsets In order to enable users to truly get rid of their mobile phones when traveling, wireless card headsets must have the function of allowing users to know incoming calls and other notification information.
  • the embodiment creatively embeds the eSIM module in the headset of the headset body.
  • the earphone control module receives the embedded instructions passed in by the recognition module, and sends instructions to the sound unit according to the following priority order, specifically,
  • the headset controls The module enters the waiting state until it receives the embedded instruction to answer the call or the embedded instruction to hang up the call from the recognition module;
  • the headset control module sends an instruction to answer the call to the eSIM module, and sends the information to play the call voice to the sound unit, and then continues to enter the waiting state; the eSIM module chooses to answer the call, and The controller makes the headset start to transmit the call voice;
  • the headset control module sends an instruction to hang up the phone to the eSIM module and an instruction to continue the previous task to the sounding unit;
  • an embedded instruction to enter the smart music mode is received in the non-smart music mode, an instruction to play music from the smart music control module is sent to the sound unit; if an embedded instruction to exit the smart music mode is received in the smart music mode , Then send an instruction to play music from the storage module to the sound unit;
  • the eSIM module has a built-in eSIM NB-IOT communication module to realize the functions of receiving incoming calls, answering calls, and hanging up calls, and transmits relevant information to the control system.
  • the earphone control module receives various embedded instructions from the recognition module, and sends instructions to the sound unit of the earphone according to the following priority order, as shown in FIG. 4:
  • the call alert information will be sent to the sounding unit first, and the sounding unit will suspend the current task and play alert music about incoming calls to the user.
  • control system After sending the incoming call warning information to the sounding unit, the control system enters a waiting state until it receives an embedded instruction to answer the call or an embedded instruction to hang up from the identification module.
  • the control system sends an instruction to answer the phone to the eSIM module, and sends a message to the sound unit to play the call voice, and then continues to enter the waiting state.
  • the eSIM module chooses to answer the call, and the controller makes the headset start to transmit the call voice.
  • control system If an embedded instruction to hang up the phone is received in the waiting state, the control system sends an instruction to hang up the phone to the eSIM module and an instruction to continue the previous task to the controller.
  • a microcontroller STC89C52RC is preferably used.
  • the controller is a CMOS 8-bit microcontroller with the following standard functions: 8k bytes of Flash, 512 bytes of RAM, 32-bit I/O ports, watchdog timer, built-in 4KB EEPROM, MAX810 reset circuit, 3 16 Bit timer/counter, 4 external interrupts, a 7-vector 4-level interrupt structure (compatible with the traditional 51 5-vector 2-level interrupt structure), full-duplex serial port.
  • An implementation system of an EEG-based smart sports headset using any one of the above-mentioned implementation methods of an EEG-based smart sports headset including an EEG signal acquisition module, an EEG signal preprocessing module, and an identification module , ESIM module and storage module,
  • EEG signal acquisition module for EEG signal acquisition, using dry electrode EEG acquisition technology, the dry electrode sensors 1a, 1b, 1c, 1d collected by the four single data channels in the EEG signal acquisition module are set by Sampling frequency for EEG signal collection;
  • EEG signal preprocessing module preprocess the collected EEG signals, the EEG signal preprocessing module receives the EEG signals collected by the EEG acquisition module, and the EEG signal preprocessing module filters and amplifies the EEG signals ;
  • Recognition module Obtain the classification results through the EEG feature extraction model, and use the EEG feature extraction method combining independent component analysis ICA and wavelet analysis WY to extract the features of the EEG signal preprocessed in step S2 to obtain the classification results; establish the brain Electrical feature deep learning model: use the back propagation algorithm to train the feedforward neural network learning model; recognize the EEG signals collected in real time, and directly preset the parameters obtained in steps S3 and S4 in the recognition module to determine the movement process of the athlete
  • the EEG signal data collected in real time in the computer is preprocessed by the preprocessing module, and then input to the recognition module to obtain the corresponding classification result; send instructions to the headset control module, and the classification result obtained after the processor of the recognition module is calculated, and The classification results in the storage module are compared, the corresponding embedded instructions are obtained, and the embedded instructions are sent to the headset control module;
  • eSIM module call control through the eSIM module to realize the control of receiving incoming calls, answering calls, and hanging up calls, and send relevant control information to the headset control module;
  • Storage module used to store the classification results of EEG signals and the corresponding embedded instructions.
  • the system of the embodiment uses a combination of wavelet analysis and independent component analysis for feature extraction, and uses a back propagation algorithm for parameter training of the neural network model, which has high accuracy.
  • the system of the embodiment obtains neural network parameters by training a large amount of labeled EEG data in advance, and presets them in the recognition module.
  • the recognition module is directly substituted into the calculation, which can save a lot of time and reduce the lag of brainwave manipulation of the headset.
  • the recognition module After obtaining the classification result in the system of the embodiment, the recognition module compares it with the classification result in the storage module, directly obtains the corresponding embedded instruction, and sends it to the headset control module to reduce the intermediate process time.
  • the system of the embodiment embeds the eSIM module in the wireless earphone, and through the earphone control module of its own design, it realizes the use of brain electricity to control and answer important calls without carrying the mobile phone, so that the user can completely get rid of the fetters of the mobile phone during exercise.
  • the implementation system of the smart sports earphone based on the brain electricity idea of the embodiment has the function of answering and hanging up the call without being connected to the mobile phone, and a smart music that can change the frequency and volume of the music according to the rhythm of people's exercise Mode, can help people solve the worries of disturbing the original exercise rhythm due to manual switching or pausing of songs during exercise, and the inconvenience of exercise caused by the need to carry mobile phones because of the need to listen to songs and answer important calls. Smart music helps people increase the sense of movement rhythm and enhance the effect of sports.
  • the function of the EEG signal acquisition module is to use four dry electrodes to collect the EEG data of the athlete's brain in real time, and send the original EEG signal to the EEG signal processing module through the USB interface.
  • the subject wears the headset and sits in front of the display screen, and performs corresponding brain imagination activities according to the instructions issued by the screen.
  • the duration is T
  • the sampling frequency is 128 Hz
  • each channel produces T ⁇ 128 data points.
  • the same task needs to be repeatedly executed n times. To ensure the accuracy and universality of the EEG database, n should be as large as possible.
  • the function of the EEG signal preprocessing module is to filter and amplify the received original EEG signal, and then send the preprocessed data to the recognition module through the USB interface.
  • the function of the recognition module is to use the preset parameters to classify and recognize the brain waves processed by the preprocessing module, and determine whether it belongs to start music, pause music, stop music, switch to the next song, switch to the previous song, answer the call, Hang up the phone, enter smart music mode, exit smart music mode, or none of them. If it belongs to one of the above nine behaviors, the classification result is compared with the classification result stored in the storage module to obtain the corresponding embedded instruction. And send the instruction to the headset control module via Bluetooth. If it does not belong to one of the above behaviors, no instructions will be issued.
  • the headphone control module of the embodiment is a control system specially designed based on brain electricity idea control, which effectively realizes the use of brain electricity idea to start music, pause music, stop music, switch the previous song, switch the next 5 earphones Control of basic functions.
  • the control system also realizes the control of answering and hanging up calls with the first priority using EEG.
  • the control system also realizes the use of brain electricity ideas to freely enter and exit the intelligent music mode.
  • the intelligent music mode of the intelligent music control module can adjust the rhythm and volume of the music in real time according to the athlete's current exercise rhythm and intensity, so that the music is no longer the same as in the past. In order to hinder the effect, it can match the exercise rhythm of the athlete, increase the rhythm and effect of the exercise of the athlete, and better assist the exercise of the athlete.
  • the embodiment creatively combines heart rate detection and brain wave detection to comprehensively measure the athlete's real-time exercise intensity, which will effectively slow down the steepness of the exerciser's exercise rhythm measurement value and make the result more accurate.
  • the heart rate measurement requires the use of a heart rate sensor.
  • the heart rate sensor is placed inside the earphone on the top of the athlete's head to detect the athlete's heart rate in real time during exercise, and transmit the heart rate data to the intelligent music control module through the USB interface in real time.
  • the intelligent music control module After receiving the real-time heart rate data from the heart rate sensor, the intelligent music control module performs real-time analysis of the exercise intensity represented by the current heart rate.
  • the relationship between real-time heart rate and resting heart rate, maximum heart rate and exercise intensity in aerobic exercise is:
  • R t (R m -R q -Y) ⁇ Q R -R q
  • R t is the real-time heart rate of aerobic exercise
  • R m is the maximum heart rate of aerobic exercise, in this formula it is 210 beats per minute
  • R q is the resting heart rate
  • Y is the age of the exerciser
  • Q R is the reason The real-time exercise intensity of the athlete measured by the heart rate.
  • the relationship between the exercise intensity of the exerciser and the real-time heart rate of the exerciser can be obtained as:
  • Q R below 0.5 means low exercise intensity
  • 0.5 to 0.75 means medium exercise intensity
  • above 0.75 means high exercise intensity
  • the brain waves used by the smart music control module are passed in by the preprocessing module.
  • the slow wave ( ⁇ , ⁇ wave) frequency of brain wave The sum is ⁇ + ⁇ , and the sum of fast wave ( ⁇ , ⁇ waves) frequency is ⁇ + ⁇ .
  • the relative power of brain waves that is, the ratio of the sum of slow wave frequencies to the sum of fast wave frequencies will increase.
  • the embodiment uses the growth rate of the relative power of the brain waves to measure the exercise intensity of the athlete. Its definition is as follows:
  • P is the relative power of the brain wave when the athlete is at rest
  • ⁇ P is the change in the relative power of the brain wave when the athlete is exercising.
  • f represents the real-time music frequency
  • v represents the real-time music volume
  • the real-time music frequency and volume can be derived as:
  • f 0 represents the initial music frequency
  • v 0 represents the initial music volume
  • the above-mentioned music frequency and music volume realize the characteristics of changing with the movement rhythm of the athletes, but they are more smooth and pleasant to listen to, which can effectively keep the athletes in the rhythm of the movement, which can bring better experience to the athletes and help the athletes Increase exercise effect.
  • the smart sports earphones based on EEG idea of the embodiment realize the integrated services of EEG signal collection, analysis and processing, classification result conversion, wireless earphone control, call answering, and smart music, and finally realize EEG idea-based services Smart sports headphones.
  • the method and system of the embodiment use a large amount of EEG data for feature extraction, classification and recognition in advance, and use these data to train the EEG neural network model, and the trained parameters Enter it into the recognition module in advance.
  • the feature extraction method adopts a combination of wavelet analysis and independent component analysis
  • the deep learning model adopts a back-propagation feedforward neural network, which makes the classification accuracy better.
  • the eSIM module has a built-in eSIM NB-IOT communication module, which can realize the functions of receiving incoming calls, answering calls, and hanging up calls, and transmits related information to the headset control module.
  • the function of the earphone control module is to receive various embedded commands sent by the recognition module, and execute various commands in a certain priority order to realize the control of various functions by the brain electricity idea.
  • the intelligent music control module receives the EEG information of the preprocessing module and the heart rate information of the heart rate sensor, comprehensively measures the athlete's current exercise rhythm and intensity, and makes corresponding changes to the frequency and volume of the music.
  • the storage module uses an SLC flash memory controlled by IS903.
  • the storage module supports the insertion of an SD card to store the songs downloaded by the athletes, which is convenient for the athletes to listen to the songs without carrying a mobile phone and connecting to Bluetooth.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Signal Processing (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Otolaryngology (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

本发明提供一种基于脑电意念的智能运动耳机及其实现方法与系统，该运动耳机包括电源和耳机本体，耳机本体包括头戴、耳机控制模块和发声单元，头戴的两端分别设有发声单元，耳机控制模块内置在头戴中，还包括脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块，脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块分别内置在头戴中，脑电信号采集模块包括干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d；该种基于脑电意念的智能运动耳机，能够通过对脑电信号的采集、预处理，进而识别后实现对耳机本体的控制，结构设计合理，便于使用。

Description

基于脑电意念的智能运动耳机及其实现方法与系统 技术领域
本发明涉及一种基于脑电意念的智能运动耳机及其实现方法与系统，基于脑电波信号采集、脑电数据深度学习模型设计和嵌入式软硬件设计的实现方案，旨在实现基于脑电意念和深度学习的智能运动耳机，属于脑电波技术和嵌入式系统的交叉领域。
背景技术
目前，随着生活水平的提高，越来越多的人选择在早晨或夜间跑步锻炼身体。而跑步时听歌，有助于人们放松和保持节奏感。跑步时听歌的人越来越多，对于能够在跑步时佩戴的耳机的需求也越来越高。目前市场上的耳机均为有线耳机或无线蓝牙耳机。然而，即便是无线蓝牙耳机，在人们跑步时需要暂停歌曲、停止音乐或者是要切歌的时候，都需要人们手动操作，而这种操作，无疑会打乱人们原有的运动节奏，使人们不得不重新花费一段时间进行调整。若是正在进行计时跑步，这种对节奏的打乱更会使得计时结果不精准，导致锻炼计划的效果不佳。因此，一种在不打乱人们运动节奏的前提下便可以实现灵活操作和控制的耳机，对人们来说十分便利。
此外，对于人们来说，运动的一个苦恼是因担心错过重要信息，而在运动时不得不携带手机在身上。这导致了人们在运动时需时刻关注手机在不在身边，有没有因运动剧烈而被甩出口袋等问题。人们也因此不敢做过于剧烈的运动。而目前世界上，还没有一款耳机能够使人们不携带手机便能听音乐且不错过任何重要的信息。
除此之外，人们运动节奏控制的好坏，对运动的效果有十分显著的影响。然而，目前还没有一款智能穿戴设备能够实现帮助人们控制运动的节奏。辅助人们控制运动节奏的设备，可以大大增加人们运动的效益。
发明内容
本发明的目的是提供一种基于脑电意念的智能运动耳机及其实现方法与系统，实现对耳机的控制，解决现有技术中存在的在运动时因听歌需要手动控制的问题。
本发明的技术解决方案是：
一种基于脑电意念的智能运动耳机，包括电源和耳机本体，耳机本体包括 头戴、耳机控制模块和发声单元，头戴的两端分别设有发声单元，耳机控制模块内置在头戴中，还包括脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块，脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块分别内置在头戴中，脑电信号采集模块包括干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d，干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d分别设于头戴上，脑电信号采集模块通过脑电信号预处理模块连接识别模块，识别模块连接储存模块。
进一步地，还包括eSIM模块，eSIM模块内置在头戴中，eSIM模块与耳机控制模块连接，eSIM模块采用eSIM NB-IOT通信模组M5310，识别模块采用STM32F072RB芯片。
进一步地，干态电极传感器1a与干态电极传感器1d、干态电极传感器1b与干态电极传感器1c分别双极并联，干态电极传感器1a与干态电极传感器1d分别设于头戴的两端，干态电极传感器1b与干态电极传感器1c分别设于头戴的两端且设于干态电极传感器1a与干态电极传感器1d之间；干态电极传感器1a与干态电极传感器1d、干态电极传感器1b与干态电极传感器1c均采用主动型干态电极传感器。
进一步地，识别模块采用STM32F072RB芯片。
一种采用上述任一项所述的基于脑电意念的智能运动耳机的实现方法，包括以下步骤，
S1、进行脑电信号采集，采用干电极脑电采集技术，由脑电信号采集模块中的四个单数据通道采集的干态电极传感器1a、1b、1c、1d通过设定的采样频率进行脑电信号采集；
S2、对采集的脑电信号进行预处理，由脑电信号预处理模块接收脑电采集模块采集到的脑电信号，脑电信号预处理模块进行脑电信号的滤波与放大；
S3、通过脑电特征提取模型获得分类结果，对步骤S2预处理后的脑电信号，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果；
S4、采用反向传播算法将步骤S3中所得分类结果带入前馈神经网络训练，求取参数隐含层神经元数n、权重矩阵W，得到本方法使用的基于前馈神经网络的脑电特征深度学习模型；
S5、对实时采集的脑电信号进行识别，将步骤S4得到的参数直接预设在识别模块中，即，将S4建立的脑电特征深度学习模型预置在识别模块中，将运动者运动过程中实时采集到的脑电信号数据，通过预处理模块预处理后，输入识别模块后，得到相应的分类结果；
S6、对耳机控制模块发送指令，经识别模块的处理器运算后得到的分类结果，与存储模块中的分类结果进行比对，得到对应的嵌入式指令，将嵌入式指令发送给耳机控制模块。
进一步地，还包括步骤S7，通过eSIM模块进行通话控制，实现收到来电、接听电话、挂断电话的控制，并将相关控制信息发送给耳机控制模块。
进一步地，步骤S3中，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果，具体为，
S31、对步骤S2预处理后的脑电信号进行二进小波变换，脑波信号的二进小波变换的定义式为：
Figure PCTCN2019092880-appb-000001
其中，c i，j二进小波系数，k为整数，
Figure PCTCN2019092880-appb-000002
为小波函数，
Figure PCTCN2019092880-appb-000003
为二进小波函数，其逆变换式为：
Figure PCTCN2019092880-appb-000004
其中，x j(t)表示信号x(t)在某一刻度(2 j)下的分量；
对于从脑电信号采集模块测得的2导脑电数据，对每导数据x i进行正交小波分解；采用长度为4的Daubechis小波，以层数为4，按如上公式分解，每导数据被分解为5个子带信号：
Figure PCTCN2019092880-appb-000005
原信号与每个子带的频率分布关系为：
Figure PCTCN2019092880-appb-000006
Figure PCTCN2019092880-appb-000007
其中，f s为采样频率；
S32、选取子带，具体为，根据需要选择同一尺度或多个尺度下的各导脑电多个子带的组合，来组成独立分量分析的输入：
Figure PCTCN2019092880-appb-000008
其中，
Figure PCTCN2019092880-appb-000009
式中，n、m为所选择的子带范围；
脑电信号有5个主要频带：Delta(0-4Hz)、Theta(4-8Hz)、Alpha(8-15Hz)、Beta(15-30Hz)和Gamma(30-60Hz)，由上述子带信号的频率范围公式，得到五个脑电频带对应的子带：Delta：
Figure PCTCN2019092880-appb-000010
Theta：
Figure PCTCN2019092880-appb-000011
Alpha：
Figure PCTCN2019092880-appb-000012
Beta：
Figure PCTCN2019092880-appb-000013
Gamma：
Figure PCTCN2019092880-appb-000014
为了增强5个脑电频带特征，选择n＝1，m＝5，即：
Figure PCTCN2019092880-appb-000015
将上述结果作为独立分量分析的输入；
S33、通过扩展Infomax算法求取分离矩阵，具体为，独立分量分析即从n个通道获得n个观测信号x i，每个观测信号是n个独立源信号s i的线性混合，即：x＝As，其中，x＝(x 1，x 2，...x n) T，s＝(s 1，s 2，...s n) T，A为常数且A不为0；ICA通过分离矩阵W将独立分量将多通道信源中分离出来，即：u＝Wx，其中，u是对真实信源s的逼近；考虑到脑电信号同时包含超高斯和亚高斯分布的独立分量成分，采用基于信息极大准则的扩展Infomax算法进行分离矩阵估计，扩展最大熵算法的解混矩阵调节公式为：
Figure PCTCN2019092880-appb-000016
其中，μ为学习步长，依具体情况进行自适应调整，I为单位矩阵，K为对角矩阵，对角元素如下：
k ii＝1(超高斯信号)
k ii＝-1(亚高斯信号)
故由递推公式W(k+1)＝W(k)+μΔW可求得矩阵W。
S34、求解独立分量，具体为，将矩阵W代回上述公式，得到独立分量u＝(u 1，u 2，...u n) T。
进一步地，步骤S4中，采用反向传播算法训练前馈神经网络学习模型，具体为：
S41、建立前向传导神经网络，采用单极性Sigmoid函数作为神经网络的激活函数，即：
Figure PCTCN2019092880-appb-000017
其中，x取值范围为(0,1)，将一个实数映射到(0,1)的区间，用来做二分类；
S42、计算各层激活值，具体为，L 2，L 3，...，L n层各神经元的激活值为：
Figure PCTCN2019092880-appb-000018
其中，
Figure PCTCN2019092880-appb-000019
表示第n-1层第j个神经元对第n层第i个神经元的权重，
Figure PCTCN2019092880-appb-000020
表示第n-1层第j个神经元的输入，b n-1表示第n-1层的偏置；
S43、计算输出层残差，具体为，向前传播的步骤如下：z n＝W n-1f(z n-1)+b n-1
其后进行反向传播算法计算，定义残差为：
Figure PCTCN2019092880-appb-000021
其中，z l为第l输入，W为第l层至第l+1层权重矩阵，b为第l层至第l+1层偏置，x为第l层各神经元的值，y为第l+1层各神经元的值，J为雅可比矩阵；
那么对于输出层的n层，残差的计算公式如下：
Figure PCTCN2019092880-appb-000022
其中，⊙为哈密顿积，a n表示第n层的输出；
S44、计算其余各层残差，具体为，对于其它n＝n 1，n 2，…n i网络层，残差按如下公式计算：
Figure PCTCN2019092880-appb-000023
S45、计算最终的偏导数值：
Figure PCTCN2019092880-appb-000024
Figure PCTCN2019092880-appb-000025
利用上述偏导值修正梯度，包括权重矩阵梯度和偏置梯度，提高神经网络的识别率；修正后的权重矩阵梯度为：
Figure PCTCN2019092880-appb-000026
式中，ΔW′ n为修正后的第n层权重矩阵梯度，ΔW n为修正前的第n层权重矩阵梯度；
修正后的偏置梯度为：
Figure PCTCN2019092880-appb-000027
式中，Δb′为修正后的第n层偏置梯度，Δb n为修正前的第n层偏置梯度；
S46、隐层神经元的数目按照经验公式选择：
Figure PCTCN2019092880-appb-000028
其中，n为隐层神经元数，n i为输入神经元数，n 0为输出神经元数，a为介于1～10之间的常数。
S47、将步骤S3中提取到的独立分量，加入标签，按上述步骤输入反向传播的前馈神经网络模型中，即可训练得到网络模型中的各参数。
进一步地，耳机控制模块中，耳机控制模块接收识别模块传入的嵌入式指令，并按照如下优先级顺序，向发声单元发出指令，具体为，
若接收到eSIM模块传入的来电信息，则优先向发声单元发送来电警示信息，并使发声单元暂停当前任务，向用户播放有来电的警示音乐；在向发声单元发送来电警示信息后，耳机控制模块进入等待状态，直至接收到识别模块传入的接听电话的嵌入式指令或挂断电话的嵌入式指令；
若在等待状态下接收到接听电话的嵌入式指令，则耳机控制模块向eSIM 模块发送接听电话的指令，向发声单元发送播放通话语音的信息，随后继续进入等待状态；eSIM模块选择接听电话，而控制器则使耳机开始传输通话语音；
若在等待状态下接收到挂断电话的嵌入式指令，则耳机控制模块向eSIM模块发送挂断电话的指令，向发声单元发送继续先前任务的指令；
若在非智能音乐模式下接收到进入智能音乐模式的嵌入式指令，则向发声单元发送播放来自智能音乐控制模块的音乐的指令；若在智能音乐模式下接收到退出智能音乐模式的嵌入式指令，则向发声单元发送播放来自存储模块的音乐的指令；
若接收到开始音乐、暂停音乐、停止音乐、切换下一首、切换上一首的嵌入式指令，则向发声单元发送相应指令。
一种采用上述任一项所述的基于脑电意念的智能运动耳机的实现方法的基于脑电意念的智能运动耳机的实现系统，包括脑电信号采集模块、脑电信号预处理模块、识别模块、eSIM模块和储存模块，
脑电信号采集模块：进行脑电信号采集，采用干电极脑电采集技术，由脑电信号采集模块中的四个单数据通道采集的干态电极传感器1a、1b、1c、1d通过设定的采样频率进行脑电信号采集；
脑电信号预处理模块：对采集的脑电信号进行预处理，由脑电信号预处理模块接收脑电采集模块采集到的脑电信号，脑电信号预处理模块进行脑电信号的滤波与放大；
识别模块：通过脑电特征提取模型获得分类结果，对步骤S2预处理后的脑电信号，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果；建立脑电特征深度学习模型：采用反向传播算法训练前馈神经网络学习模型；对实时采集的脑电信号进行识别，将步骤S3、S4得到的参数直接预设在识别模块中，将运动者运动过程中实时采集到的脑电信号数据，通过预处理模块预处理后，输入识别模块后，得到相应的分类结果；对耳机控制模块发送指令，经识别模块的处理器运算后得到的分类结果，与存储模块中的分类结果进行比对，得到对应的嵌入式指令，将嵌入式指令发送给耳机控制模块；
eSIM模块：通过eSIM模块进行通话控制，实现收到来电、接听电话、挂断电话的控制，并将相关控制信息发送给耳机控制模块；
存储模块：用于存储脑电信号的分类结果以及与其对应的嵌入式指令。
进一步地，还包括智能音乐控制模块：接收心率传感器采集的实时心率数据，实现音乐的音量和频率随运动者的实时心率数据的变化而进行相应变化，具体为，
接收到心率传感器传入的实时心率数据后，对当前心率所代表的运动强度进行实时分析，运动者的运动强度与运动者的实时心率的关系为：
Figure PCTCN2019092880-appb-000029
Figure PCTCN2019092880-appb-000030
式中，R t为有氧运动的实时心率，R m为有氧运动的最大心率，R q为安静心率，Y为运动者的年龄，Q R为由心率测得的运动者的实时运动强度；
除使用心率衡量运动强度外，还需使用脑电波衡量，使用脑电波相对功率的增长率来衡量运动者的运动强度，其定义如下：
Figure PCTCN2019092880-appb-000031
式中，P为运动者安静时的脑电波相对功率，ΔP为运动者运动时脑电波相对功率的变化量；
使用如下公式最终确定音乐频率和音量的变化情况，实时音乐频率和音量为：
Figure PCTCN2019092880-appb-000032
Figure PCTCN2019092880-appb-000033
式中，f 0表示初始音乐频率，v 0表示初始音乐音量。
本发明的有益效果是：
一、该种基于脑电意念的智能运动耳机，能够通过对脑电信号的采集、预处理，进而识别后实现对耳机本体的控制，结构设计合理，便于使用。通过脑电信号采集模块实现对脑电信号的采集，并通过脑电信号预处理模块实现对采集脑电信号的预处理，通过识别模块实现对脑电信号的识别，进而发送控制指令给耳机控制模块，实现对发声单元的控制。该种基于脑电意念的智能运动耳机，创造性地将eSIM模组内置于无线耳机中，通过所属设计耳机控制模块，实现了在不携带手机的情况下，用脑电意念控制接听重要来电，使用户在运动时彻底摆脱手机的羁绊。
二、该种基于脑电意念的智能运动耳机及其实现方法与系统，将特征提取、神经网络各项参数预先输入至处理模块中，省去传统脑电处理中的预处理、特 征提取等步骤，能够快速、准确的获得脑电特征并进行分类。
三、本发明采用小波分析与独立分量分析相结合的特征提取方法，采用反向传播的前馈神经网络，使分类结果更精准。
四、本发明实现了对脑电信号的采集、分析处理、分类结果转化、无线耳机控制、来电接听、智能音乐的一体式服务，最终实现了基于脑电意念的智能运动耳机。
五、该种基于脑电意念的智能运动耳机及其实现方法与系统，创造性地将脑电波技术和无线插卡耳机技术的控制结合在一起，实现一种全新的耳机控制方式，能够很好地帮助运动者摆脱因需要手动控制耳机而打乱运动节奏的烦恼问题。
附图说明
图1是实施例中耳机本体的结构示意图。
图2是本发明实施例基于脑电意念的智能运动耳机实现系统的说明框图；
图3是实施例中特征提取、分类识别参数确定流程示意图。
图4是实施例中耳机控制模块的ASM图。
具体实施方式
下面结合附图详细说明本发明的优选实施例。
实施例
一种基于脑电意念的智能运动耳机，如图1，包括电源和耳机本体，耳机本体包括头戴、耳机控制模块和发声单元，头戴的两端分别设有发声单元，耳机控制模块内置在头戴中，还包括脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块，脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块分别内置在头戴中，脑电信号采集模块包括干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d，干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d分别设于头戴上，脑电信号采集模块通过脑电信号预处理模块连接识别模块，识别模块连接储存模块。
该种基于脑电意念的智能运动耳机，能够实现对耳机本体的控制，通过脑电信号采集模块实现对脑电信号的采集，并通过脑电信号预处理模块实现对采集脑电信号的预处理，通过识别模块实现对脑电信号的识别，进而发送控制指 令给耳机控制模块，实现对发声单元的控制。该种基于脑电意念的智能运动耳机，创造性地将eSIM模组内置于无线耳机中，通过所属设计耳机控制模块，实现了在不携带手机的情况下，用脑电意念控制接听重要来电，使用户在运动时彻底摆脱手机的羁绊。
该种基于脑电意念的智能运动耳机，还包括eSIM模块，eSIM模块内置在头戴中，eSIM模块与耳机控制模块连接，eSIM模块采用eSIM NB-IOT通信模组M5310，识别模块采用STM32F072RB芯片。为减少耳机的重量和体积，增加耳机的便携度和信号强度，实施例优选采用由中国移动公司发布的eSIM NB-IOT通信模组M5310。该模组拥有世界上同类产品中较小的尺寸，仅19×18.4×2.7mm，节省布板面积达30％以上。模组采用的是海思Hi2110芯片，支持eSIM技术以及OneNET平台协议。
识别模块采用STM32F072RB芯片。为提高运算速度，减少运动者运动想象至耳机反映的迟滞时间，识别模块采用STM32F072RB芯片。STM32F072RB芯片是基于ARM Cortex-M7的32位微处理器。其拥有一个64位AXI系统总线接口、一个指令和数据高速缓存、64位指令紧耦合存储器(ITCM)、双32位数据紧耦合存储器(DTCM)，具有极高的运算性能。
脑电信号采集模块中，干态电极传感器1a与干态电极传感器1d、干态电极传感器1b与干态电极传感器1c分别双极并联，干态电极传感器1a与干态电极传感器1d分别设于头戴的两端，干态电极传感器1b与干态电极传感器1c分别设于头戴的两端且设于干态电极传感器1a与干态电极传感器1d之间；干态电极传感器1a与干态电极传感器1d、干态电极传感器1b与干态电极传感器1c均采用主动型干态电极传感器。
传统的脑电波数据采集技术通常采用湿电极采集技术，被测试者在测试前，通常需要涂抹导电介质。运动人员在运动过程中通常头部会出现汗水，采用传统的脑电波数据采集技术，会影响数据的准确性。另外，在每次运动前涂抹导电介质，会大大增加每次运动的繁琐度。因此，实施例的脑电信号采集模块，脑电波的数据采集采用干电极脑电采集技术。相比传统脑电数据采集技术，干电极采集技术借助微针技术、超高输入阻抗放大器等技术，摆脱对导电介质的依赖，具有简便易使用性。
为保证数据的稳定性，实施例的脑电信号采集模块由四个干态电极传感器 1a、1b、1c、1d构成。1a与1d采用双极并联的方式，分别位于无线插卡耳机上方的头皮区域的两侧。1b与1c同样采用双极并联的方式，分别位于预处理模块和存储模块的下方。脑电信号采集模块的采样频率为128Hz。采集到的2路脑电信号分别为两组活动电极之间的电位差。干电极传感器采用由g.tec公司生产的g.SAHARA型号的主动型干电极。
一种采用上述任一项所述的基于脑电意念的智能运动耳机的实现方法，如图2，包括以下步骤，
S1、采集：进行脑电信号采集，采用干电极脑电采集技术，由脑电信号采集模块中的四个单数据通道采集的干态电极传感器1a、1b、1c、1d通过设定的采样频率进行脑电信号采集。
S2、预处理：对采集的脑电信号进行预处理，由脑电信号预处理模块接收脑电采集模块采集到的脑电信号，脑电信号预处理模块进行脑电信号的滤波与放大；
步骤S1采集到的2导脑电信号，通过USB接口，发送到脑电信号预处理模块。脑电信号预处理模块进行脑电信号的滤波与放大。脑电信号预处理模块依次包含第一级放大器、50Hz陷波器、低通滤波器、带通滤波器、第二级放大器。
S3、分类：通过脑电特征提取模型获得分类结果，对步骤S2预处理后的脑电信号，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果；
步骤S3中，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果，如图3，具体为，
S31、对步骤S2预处理后的脑电信号进行二进小波变换，脑波信号的二进小波变换的定义式为：
Figure PCTCN2019092880-appb-000034
其逆变换式为：
Figure PCTCN2019092880-appb-000035
其中，x j(t)表示信号x(t)在某一刻度(2 j)下的分量；
对于从脑电信号采集模块测得的2导脑电数据，对每导数据x i进行正交小波分解；采用长度为4的Daubechis小波，以层数为4，按如上公式分解，每导数据被分解为5个子带信号：
Figure PCTCN2019092880-appb-000036
原信号与每个子带的频率分布关系为：
Figure PCTCN2019092880-appb-000037
Figure PCTCN2019092880-appb-000038
其中，f s为采样频率；
S32、选取子带，具体为，根据需要选择同一尺度或多个尺度下的各导脑电多个子带的组合，来组成独立分量分析的输入：
Figure PCTCN2019092880-appb-000039
其中，
Figure PCTCN2019092880-appb-000040
式中，n、m为所选择的子带范围；
脑电信号有5个主要频带：Delta(0-4Hz)、Theta(4-8Hz)、Alpha(8-15Hz)、Beta(15-30Hz)和Gamma(30-60Hz)，由上述子带信号的频率范围公式，得到五个脑电频带对应的子带：Delta：
Figure PCTCN2019092880-appb-000041
Theta：
Figure PCTCN2019092880-appb-000042
Alpha：
Figure PCTCN2019092880-appb-000043
Beta：
Figure PCTCN2019092880-appb-000044
Gamma：
Figure PCTCN2019092880-appb-000045
为了增强5个脑电频带特征，选择n＝1，m＝5，即：
Figure PCTCN2019092880-appb-000046
将上述结果作为独立分量分析的输入；
S33、通过扩展Infomax算法求取分离矩阵，具体为，独立分量分析即从n个通道获得n个观测信号x i，每个观测信号是n个独立源信号s i的线性混合，即：x＝As，其中，x＝(x 1，x 2，...x n) T，s＝(s 1，s 2，...s n) T；ICA通过分离矩阵W将独立分量将多通道信源中分离出来，即：u＝Wx，其中，u是对真实信源s的逼近；考虑到脑电信号同时包含超高斯和亚高斯分布的独立分量成分，采用基于信息极大准则的扩展Infomax算法进行分离矩阵估计。扩展Infomax算法，即扩展最大熵算法，是根据自然梯度法求解分离矩阵W。其目标判据为独立分量通过非线性环节后目标熵极大。该算法能够同时适用于超高斯和亚高斯信号，而且避免了矩阵求逆的问题，使收敛速度得到提升。
扩展最大熵算法的解混矩阵调节公式为：
Figure PCTCN2019092880-appb-000047
Figure PCTCN2019092880-appb-000048
其中，I为单位矩阵，K为对角矩阵，对角元素如下：
k ii＝1(超高斯信号)
k ii＝-1(亚高斯信号)
S34、求解独立分量，具体为，将矩阵W代回上述公式，得到独立分量u＝(u 1，u 2，...u n) T。
独立分量分析(ICA)的优势在于收敛精度高和速度快。与主成分分析(PCA)等传统特征分析方法相比，ICA是基于高阶统计特性的分析方法。在很多应用中，对高阶统计特性的分析更符合实际。小波分析(WY)的优势在于分析非平稳信号。它通过时频窗口对信号进行分解，能够增强脑电信号中的 瞬态信息。实施例方法将独立分量分析与小波分析相结合，可获得更为有效的分类结果。
S4、训练：建立脑电特征深度学习模型：采用反向传播算法训练前馈神经网络学习模型；
为使神经网络训练效果更佳，神经网络每一层的权重参数和偏置参数的获取更迅速和合理，实施例方法采用反向传播算法训练前馈神经网络学习模型。
步骤S4中，采用反向传播算法训练前馈神经网络学习模型，具体为：
S41、建立前向传导神经网络，为了增加系统的非线性，采用单极性Sigmoid函数作为神经网络的激活函数，即：
Figure PCTCN2019092880-appb-000049
S42、计算各层激活值，具体为，L 2，L 3，...，L n层各神经元的激活值为：
Figure PCTCN2019092880-appb-000050
其中，
Figure PCTCN2019092880-appb-000051
表示第n-1层第j个神经元对第n层第i个神经元的权重，
Figure PCTCN2019092880-appb-000052
表示第n-1层第j个神经元的输入，b n-1表示第n-1层的偏置；
S43、计算输出层残差，具体为，向前传播的步骤如下：z n＝W n-1f(z n-1)+b n-1
其后进行反向传播算法计算，定义残差为：
Figure PCTCN2019092880-appb-000053
那么对于输出层(n层)，残差的计算公式如下：
Figure PCTCN2019092880-appb-000054
其中，⊙为哈密顿积，a n表示第n层的输出；
S44、计算其余各层残差，具体为，对于其它n＝n 1，n 2，…n i网络层，残差按如下公式计算：
Figure PCTCN2019092880-appb-000055
S45、计算最终的偏导数值：
Figure PCTCN2019092880-appb-000056
Figure PCTCN2019092880-appb-000057
利用上述偏导值修正梯度，包括权重矩阵梯度和偏置梯度，提高神经网络的识别率；修正后的权重矩阵梯度为：
Figure PCTCN2019092880-appb-000058
式中，ΔW′ n为修正后的第n层权重矩阵梯度，ΔW n为修正前的第n层权重矩阵梯度；
修正后的偏置梯度为：
Figure PCTCN2019092880-appb-000059
式中，Δb′为修正后的第n层偏置梯度，Δb n为修正前的第n层偏置梯度；
S46、利用上述偏导值修正梯度，提高神经网络的识别率；
隐层神经元的数目按照经验公式选择：
Figure PCTCN2019092880-appb-000060
其中，n为隐层神经元数，n i为输入神经元数，n 0为输出神经元数，a为介于1～10之间的常数。
S47、将步骤S3中提取到的独立分量，加入标签，按上述步骤输入反向传播的前馈神经网络模型中，即可训练得到网络模型中的各参数。
S5、识别：对实时采集的脑电信号进行识别，将步骤S3、S4得到的参数直接预设在识别模块中，将运动者运动过程中实时采集到的脑电信号数据，通过预处理模块预处理后，输入识别模块后，得到相应的分类结果；
S6、控制：对耳机控制模块发送指令，经识别模块的处理器运算后得到的分类结果，与存储模块中的分类结果进行比对，得到对应的嵌入式指令，将嵌入式指令发送给耳机控制模块。
实施例还包括步骤S7，通过eSIM模块进行通话控制，实现收到来电、接听电话、挂断电话的控制，并将相关控制信息发送给耳机控制模块。为了能使用户真正地在出行时摆脱手机，无线插卡耳机必须具备使用户得知来电等通知信息的功能。为了能使用户接收到来电信息，实施例创造性地将eSIM模组内嵌于耳机本体的头戴中。
实施例的耳机控制模块中，耳机控制模块接收识别模块传入的嵌入式指令，并按照如下优先级顺序，向发声单元发出指令，具体为，
若接收到eSIM模块传入的来电信息，则优先向发声单元发送来电警示信息，并使发声单元暂停当前任务，向用户播放有来电的警示音乐；在向发声单元发送来电警示信息后，耳机控制模块进入等待状态，直至接收到识别模块传入的接听电话的嵌入式指令或挂断电话的嵌入式指令；
若在等待状态下接收到接听电话的嵌入式指令，则耳机控制模块向eSIM 模块发送接听电话的指令，向发声单元发送播放通话语音的信息，随后继续进入等待状态；eSIM模块选择接听电话，而控制器则使耳机开始传输通话语音；
若在等待状态下接收到挂断电话的嵌入式指令，则耳机控制模块向eSIM模块发送挂断电话的指令，向发声单元发送继续先前任务的指令；
若在非智能音乐模式下接收到进入智能音乐模式的嵌入式指令，则向发声单元发送播放来自智能音乐控制模块的音乐的指令；若在智能音乐模式下接收到退出智能音乐模式的嵌入式指令，则向发声单元发送播放来自存储模块的音乐的指令；
若接收到开始音乐、暂停音乐、停止音乐、切换下一首、切换上一首的嵌入式指令，则向发声单元发送相应指令。
实施例中，eSIM模块内置上述eSIM NB-IOT通信模组，实现收到来电、接听电话、挂断电话的功能，并将相关信息传入控制系统。
实施例中，耳机控制模块接收识别模块传入的各类嵌入式指令，并按照如下的优先级顺序，如图4，向耳机的发声单元发出指令：
1)若接收到eSIM模块传入的来电信息，则优先向发声单元发送来电警示信息，并使发声单元暂停当前任务，向用户播放有来电的警示音乐。
2)在向发声单元发送来电警示信息后，控制系统进入等待状态，直至接收到识别模块传入的接听电话的嵌入式指令或挂断电话的嵌入式指令。
3)若在等待状态下接收到接听电话的嵌入式指令，则控制系统向eSIM模块发送接听电话的指令，向发声单元发送播放通话语音的信息，随后继续进入等待状态。eSIM模块选择接听电话，而控制器则使耳机开始传输通话语音。
4)若在等待状态下接收到挂断电话的嵌入式指令，则控制系统向eSIM模块发送挂断电话的指令，向控制器发送继续先前任务的指令。
5)若在非智能音乐模式下接收到进入智能音乐模式的嵌入式指令，则向发声单元发送播放来自智能音乐控制模块的音乐的指令。
6)若在智能音乐模式下接收到退出智能音乐模式的嵌入式指令，则向发声单元发送播放来自存储模块的音乐的指令。
7)若在接收到开始音乐、暂停音乐、停止音乐、切换下一首、切换上一首的嵌入式指令，则向发声单元发送相应指令。
由以上优先级顺序生成的ASM图详见附图3。其明细表如下表格所示：
Figure PCTCN2019092880-appb-000061
由于耳机控制模块执行的指令并不复杂，为降低功耗和节省成本，优选采用微控制器STC89C52RC。该控制器是CMOS8位微控制器，具有以下标准功能：8k字节Flash，512字节RAM，32位I/O口线，看门狗定时器，内置4KB EEPROM，MAX810复位电路，3个16位定时器/计数器，4个外部中断，一个7向量4级中断结构(兼容传统51的5向量2级中断结构)，全双工串行口。
一种采用上述任一项所述的基于脑电意念的智能运动耳机的实现方法的基于脑电意念的智能运动耳机的实现系统，包括脑电信号采集模块、脑电信号预处理模块、识别模块、eSIM模块和储存模块，
脑电信号采集模块：进行脑电信号采集，采用干电极脑电采集技术，由脑电信号采集模块中的四个单数据通道采集的干态电极传感器1a、1b、1c、1d通过设定的采样频率进行脑电信号采集；
脑电信号预处理模块：对采集的脑电信号进行预处理，由脑电信号预处理模块接收脑电采集模块采集到的脑电信号，脑电信号预处理模块进行脑电信号的滤波与放大；
识别模块：通过脑电特征提取模型获得分类结果，对步骤S2预处理后的脑电信号，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果；建立脑电特征深度学习模型：采用反向传播算法训练前馈神经网络学习模型；对实时采集的脑电信号进行识别，将步骤S3、S4得到的参数直接预设在识别模块中，将运动者运动过程中实时采集到的脑电信号数据，通过预处理模块预处理后，输入识别模块后，得到相应的分类结果；对耳机控制模块发送指令，经识别模块的处理器运算后得到的分类结果，与存储模块中的分类结果进行比对，得到对应的嵌入式指令，将嵌入式指令发送给耳机控制模块；
eSIM模块：通过eSIM模块进行通话控制，实现收到来电、接听电话、挂断电话的控制，并将相关控制信息发送给耳机控制模块；
存储模块：用于存储脑电信号的分类结果以及与其对应的嵌入式指令。
实施例系统使用小波分析与独立分量分析结合的方法进行特征提取，使用反向传播算法进行神经网络模型的参数训练，具有较高的准确度。实施例系统通过提前训练大量带标签的脑电数据，获得神经网络参数，预置在识别模块中。对于运动者的实时脑电数据，直接由识别模块代入运算，可节省大量时间，减少脑电波操控耳机的迟滞感。实施例系统中识别模块在获得分类结果后，将其与存储模块中的分类结果进行对比，直接获取对应的嵌入式指令，发送给耳机控制模块，减少中间过程时间。实施例系统将eSIM模组内置于无线耳机中，通过所属设计耳机控制模块，实现了在不携带手机的情况下，用脑电意念控制 接听重要来电，使用户在运动时彻底摆脱手机的羁绊。
实施例的基于脑电意念的智能运动耳机的实现系统，具有在免于与手机连接的情况下接听来电、挂断来电的功能和一种可以随人们运动节奏而改变音乐频率与音量的智能音乐模式，可以帮助人们解决在运动时因听歌需要手动切换或暂停歌曲而打乱原有运动节奏的苦恼、因需听歌和接听重要来电而需携带手机导致运动不便的烦恼，并且还可以通过智能音乐帮助人们增加运动节奏感、增强运动效果。
实施例中，脑电信号采集模块的功能是利用四个干电极，实时采集运动者脑部的EEG数据，并通过USB接口将脑电原始信号发送至脑电信号处理模块中。在运动想象阶段，受试者佩戴该耳机坐在显示屏幕前，依据屏幕发出的指令做出相应的脑部想象活动。每次实验显示屏幕发出一个指令，受试者根据指令完成一种任务的想象，持续时间为T，采样频率为128Hz，每个通道产生T·128个数据点。在每个任务周期中，需反复执行同一个任务n次。为保证脑电数据库的准确性和广泛性，n应尽可能取最大值。任务一共由九种类型，包括开始音乐任务、暂停音乐任务、停止音乐任务、切换上一首音乐任务、切换下一首音乐任务、接听来电任务、挂断来电任务、进入智能音乐模式任务、退出智能音乐模式任务。在每一任务下，屏幕持续T时间出现开始音乐的提示，受试者根据此提示想象该任务下的音乐，记录此周期的脑电数据并设置分类结果为任务下的音乐。
脑电信号预处理模块的功能是对接收到的原始脑电波信号进行滤波和放大处理，再通过USB接口将经预处理后的数据发送至识别模块中。识别模块的功能是利用预置的参数，对经预处理模块处理后的脑电波进行分类识别，判断其属于开始音乐、暂停音乐、停止音乐、切换下一首、切换上一首、接听电话、挂断电话、进入智能音乐模式、退出智能音乐模式中的某一种或均不属于。若属于上述九种行为的某一种，则将分类结果与存储模块中存储的分类结果进行比对，得到对应的嵌入式指令。并通过蓝牙向耳机控制模块发送该指令。若不属于上述行为的某一种，则不发出任何指令。
实施例的耳机控制模块是基于脑电意念控制而专门设计的控制系统，有效地实现了利用脑电意念对耳机开始音乐、暂停音乐、停止音乐、切换上一首、切换下一首5个耳机基本功能的操控。除此之外，该控制系统还实现了利用脑 电意念以第一优先级对接听电话、挂断电话的控制。该控制系统还实现了利用脑电意念自由进入和退出智能音乐模式。
智能音乐控制模块的智能音乐模式能够根据运动者当前的运动节奏和强度，实时调节音乐的节奏和音量，使得音乐不再像过往一样，常常与运动者的运动节奏不合，对运动者的运动起到阻碍作用，而是能够与运动者的运动节奏相吻合，增加运动者运动的节奏感和效果，更好的辅助运动者的运动。
实施例创造性地将心率检测和脑电波检测结合起来，综合衡量运动者的实时运动强度，将能有效地减缓运动者运动节奏测量值的陡峭度，使结果更为精准。
对心率的测量需使用心率传感器。心率传感器置于运动者头部顶端的耳机内测，实时检测运动者运动时的心率，并将心率数据通过USB接口实时传输给智能音乐控制模块。智能音乐控制模块接收到心率传感器传入的实时心率数据后，对当前心率所代表的运动强度进行实时分析。对于普通运动者而言，有氧运动中的实时心率与安静心率、最大心率、运动强度的关系式为：
R t＝(R m-R q-Y)×Q R-R q
式中，R t为有氧运动的实时心率，R m为有氧运动的最大心率，在本式中为210次每分钟，R q为安静心率，Y为运动者的年龄，Q R为由心率测得的运动者的实时运动强度。
由上述公式可以得到运动者的运动强度与运动者的实时心率的关系为：
Figure PCTCN2019092880-appb-000062
其中，Q R在0.5以下为低运动强度，0.5～0.75为中运动强度，0.75以上为高运动强度。
除使用心率衡量运动强度外，还需使用脑电波衡量。智能音乐控制模块使用的脑电波由预处理模块传入。对于不同采样点的δ波(0.5～3Hz)、θ波(4～7Hz)、α波(8～13Hz)、β波(14～40Hz)频率，脑电波的慢波(δ、θ波)频率之和为δ+θ，快波(α、β波)频率之和为α+β。
运动者在进行大量有氧运动时，慢波功率值将显著增加，快波功率值则显著减少，脑电波的相对功率即慢波频率之和与快波频率之和的比值将会增加。实施例使用脑电波相对功率的增长率来衡量运动者的运动强度。其定义如下：
Figure PCTCN2019092880-appb-000063
式中，P为运动者安静时的脑电波相对功率，ΔP为运动者运动时脑电波相对功率的变化量。
为了能够综合利用由心率和脑电测得的运动强度值，并消除量纲的影响，且要尽量减少运动者运动强度测量值的陡峭度，增加运动强度测量值的精确度和平缓性，使音乐变化平稳而不刺耳，实施例使用如下公式最终确定音乐频率和音量的变化情况：
Figure PCTCN2019092880-appb-000064
Figure PCTCN2019092880-appb-000065
式中，f表示实时音乐频率，v表示实时音乐音量。
由上式可推得实时音乐频率和音量为：
Figure PCTCN2019092880-appb-000066
Figure PCTCN2019092880-appb-000067
式中，f 0表示初始音乐频率，v 0表示初始音乐音量。
上述音乐频率和音乐音量实现了随运动者运动节奏而变化的特征，却更加的平缓动听，可以有效地使运动者保持运动的节奏感，可以给运动者带来较好的体验，帮助运动者增加运动效果。
实施例的基于脑电意念的智能运动耳机，实现了对脑电信号的采集、分析处理、分类结果转化、无线耳机控制、来电接听、智能音乐的一体式服务，最终实现了基于脑电意念的智能运动耳机。
由于脑电信号的特征提取和分类识别比较耗时，实施例方法与系统预先利用大量脑电数据进行特征提取、分类识别，并利用这些数据进行脑电神经网络模型的训练，将训练好的参数预先输入至识别模块中。实施例中，特征提取方法采用小波分析与独立分量分析结合的方法，深度学习模型采用反向传播的前馈神经网络，使分类准确度较好。
实施例中，eSIM模块内置eSIM NB-IOT通信模组，可以实现收到来电、 接听电话、挂断电话的功能，并将相关信息传入耳机控制模块。耳机控制模块的功能是接收识别模块发送的各类嵌入式指令，并以某种优先级顺序执行各类指令，实现脑电意念对各类功能的控制。智能音乐控制模块通过接收预处理模块的脑电信息和心率传感器的心率信息，综合衡量运动者当前运动节奏和强度，并使音乐的频率和音量做出相应的改变。
实施例中，为了增加读写速度，减少反映迟滞时间，存储模块使用IS903主控的SLC闪存。另外，存储模块支持插入SD卡，用于存储运动者下载的歌曲，方便运动者在不携带手机、连接蓝牙的情况下听歌。

Claims (10)

  1. 一种基于脑电意念的智能运动耳机，包括电源和耳机本体，耳机本体包括头戴、耳机控制模块和发声单元，头戴的两端分别设有发声单元，耳机控制模块内置在头戴中，其特征在于：还包括脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块，脑电信号采集模块、脑电信号预处理模块、识别模块和储存模块分别内置在头戴中，脑电信号采集模块包括干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d，干态电极传感器1a、干态电极传感器1b、干态电极传感器1c和干态电极传感器1d分别设于头戴上，脑电信号采集模块通过脑电信号预处理模块连接识别模块，识别模块连接储存模块。
  2. 如权利要求1所述的基于脑电意念的智能运动耳机，其特征在于：还包括eSIM模块，eSIM模块内置在头戴中，eSIM模块与耳机控制模块连接，eSIM模块采用eSIM NB-IOT通信模组M5310，识别模块采用STM32F072RB芯片。
  3. 如权利要求1或2所述的基于脑电意念的智能运动耳机，其特征在于：干态电极传感器1a与干态电极传感器1d、干态电极传感器1b与干态电极传感器1c分别双极并联，干态电极传感器1a与干态电极传感器1d分别设于头戴的两端，干态电极传感器1b与干态电极传感器1c分别设于头戴的两端且设于干态电极传感器1a与干态电极传感器1d之间；干态电极传感器1a与干态电极传感器1d、干态电极传感器1b与干态电极传感器1c均采用主动型干态电极传感器。
  4. 一种采用权利要求1-3任一项所述的基于脑电意念的智能运动耳机的实现方法，其特征在于：包括以下步骤，
    S1、采集：进行脑电信号采集，采用干电极脑电采集技术，由脑电信号采集模块中的四个单数据通道采集的干态电极传感器1a、1b、1c、1d通过设定的采样频率进行脑电信号采集；
    S2、预处理：对采集的脑电信号进行预处理，由脑电信号预处理模块接收脑电采集模块采集到的脑电信号，脑电信号预处理模块进行脑电信号的滤波与放大；
    S3、分类：通过脑电特征提取模型获得分类结果，对步骤S2预处理后的脑电信号，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行 特征提取，获得分类结果；
    S4、训练：采用反向传播算法将步骤S3中所得分类结果带入前馈神经网络训练，求取参数隐含层神经元数n、权重矩阵W，得到基于前馈神经网络的脑电特征深度学习模型；
    S5、识别：对实时采集的脑电信号进行识别，将步骤S4得到的参数直接预设在识别模块中，即，将S4建立的脑电特征深度学习模型预置在识别模块中，将运动者运动过程中实时采集到的脑电信号数据，通过预处理模块预处理后，输入识别模块后，得到相应的分类结果；
    S6、控制：对耳机控制模块发送指令，经识别模块的处理器运算后得到的分类结果，与存储模块中的分类结果进行比对，得到对应的嵌入式指令，将嵌入式指令发送给耳机控制模块。
  5. 如权利要求4所述的基于脑电意念的智能运动耳机实现方法，其特征在于：还包括步骤S7，通过eSIM模块进行通话控制，实现收到来电、接听电话、挂断电话的控制，并将相关控制信息发送给耳机控制模块。
  6. 如权利要求5所述的基于脑电意念的智能运动耳机实现方法，其特征在于：步骤S3中，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果，具体为，
    S31、对步骤S2预处理后的脑电信号进行二进小波变换，脑波信号的二进小波变换的定义式为：
    Figure PCTCN2019092880-appb-100001
    其中，c i，j二进小波系数，k为整数，
    Figure PCTCN2019092880-appb-100002
    为小波函数，
    Figure PCTCN2019092880-appb-100003
    为二进小波函数，其逆变换式为：
    Figure PCTCN2019092880-appb-100004
    其中，x j(t)表示信号x(t)在某一刻度(2 j)下的分量；
    对于从脑电信号采集模块测得的2导脑电数据，对每导数据x i进行正交小波分解；采用长度为4的Daubechis小波，以层数为4，按如上公式分解，每导数据被分解为5个子带信号：
    Figure PCTCN2019092880-appb-100005
    原信号与每个子带的频率分布关系为：
    Figure PCTCN2019092880-appb-100006
    Figure PCTCN2019092880-appb-100007
    其中，f s为采样频率；
    S32、选取子带，具体为，根据需要选择同一尺度或多个尺度下的各导脑电多个子带的组合，来组成独立分量分析的输入：
    Figure PCTCN2019092880-appb-100008
    其中，
    Figure PCTCN2019092880-appb-100009
    式中，n、m为所选择的子带范围；
    脑电信号有5个主要频带：Delta(0-4Hz)、Theta(4-8Hz)、Alpha(8-15Hz)、Beta(15-30Hz)和Gamma(30-60Hz)，由上述子带信号的频率范围公式，得到五个脑电频带对应的子带：Delta：
    Figure PCTCN2019092880-appb-100010
    Theta：
    Figure PCTCN2019092880-appb-100011
    Alpha：
    Figure PCTCN2019092880-appb-100012
    Beta：
    Figure PCTCN2019092880-appb-100013
    Gamma：
    Figure PCTCN2019092880-appb-100014
    为了增强5个脑电频带特征，选择n＝1，m＝5，即：
    Figure PCTCN2019092880-appb-100015
    将上述结果作为独立分量分析的输入；
    S33、通过扩展Infomax算法求取分离矩阵，具体为，独立分量分析即从n个通道获得n个观测信号x i，每个观测信号是n个独立源信号s i的线性混合，即：x＝As，其中，x＝(x 1，x 2，…x n) T，s＝(s 1，s 2，…s n) T，A为常数且A不为0；ICA通过分离矩阵W将独立分量将多通道信源中分离出来，即：u＝Wx，其中，u是对真实信源s的逼近；考虑到脑电信号同时包含超高斯和亚高斯分布的独立分量成分，采用基于信息极大准则的扩展Infomax算法进行分离矩阵估计，扩展最大熵算法的解混矩阵调节公式为：
    Figure PCTCN2019092880-appb-100016
    其中，μ为学习步长，依具体情况进行自适应调整，I为单位矩阵，K为对角矩阵，对角元素如下：
    k ii＝1(超高斯信号)
    k ii＝-1(亚高斯信号)
    故由递推公式W(k+1)＝W(k)+μΔW可求得矩阵W。
    S34、求解独立分量，具体为，将矩阵W代回上述公式，得到独立分量u＝(u 1，u 2，…u n) T。
  7. 如权利要求6所述的基于脑电意念的智能运动耳机实现方法，其特征在于：步骤S4中，采用反向传播算法训练前馈神经网络学习模型，具体为：
    S41、建立前向传导神经网络，采用单极性Sigmoid函数作为神经网络的激活函数，即：
    Figure PCTCN2019092880-appb-100017
    其中，x取值范围为(0,1)，将一个实数映射到(0,1)的区间，用来做二分类；
    S42、计算各层激活值，具体为，L 2，L 3，...，L n层各神经元的激活值为：
    Figure PCTCN2019092880-appb-100018
    其中，
    Figure PCTCN2019092880-appb-100019
    表示第n-1层第j个神经元对第n层第i个神经元的权重，
    Figure PCTCN2019092880-appb-100020
    表示第n-1层第j个神经元的输入，b n-1表示第n-1层的偏置；
    S43、计算输出层残差，具体为，向前传播的步骤如下：z n＝W n-1f(z n-1)+b n-1
    其后进行反向传播算法计算，定义残差为：
    Figure PCTCN2019092880-appb-100021
    其中，z l为第l输入，W为第l层至第l+1层权重矩阵，b为第l层至第l+1层偏置，x为第l层各神经元的值，y为第l+1层各神经元的值，J为雅可比矩阵；
    那么对于输出层的n层，残差的计算公式如下：
    Figure PCTCN2019092880-appb-100022
    其中，⊙为哈密顿积，a n表示第n层的输出；
    S44、计算其余各层残差，具体为，对于其它n＝n 1，n 2，…n i网络层，残差按如下公式计算：
    Figure PCTCN2019092880-appb-100023
    S45、计算最终的偏导数值：
    Figure PCTCN2019092880-appb-100024
    Figure PCTCN2019092880-appb-100025
    利用上述偏导值修正梯度，包括权重矩阵梯度和偏置梯度，提高神经网络的识别率；修正后的权重矩阵梯度为：
    Figure PCTCN2019092880-appb-100026
    式中，ΔW′ n为修正后的第n层权重矩阵梯度，ΔW n为修正前的第n层权重矩阵梯度；
    修正后的偏置梯度为：
    Figure PCTCN2019092880-appb-100027
    式中，Δb′为修正后的第n层偏置梯度，Δb n为修正前的第n层偏置梯度；
    S46、隐层神经元的数目按照经验公式选择：
    Figure PCTCN2019092880-appb-100028
    其中，n为隐层神经元数，n i为输入神经元数，n 0为输出神经元数，a为介于1～10之间的常数。
    S47、将步骤S3中提取到的独立分量，加入标签，按上述步骤输入反向传播的前馈神经网络模型中，即可训练得到网络模型中的各参数。
  8. 如权利要求4-7任一项所述的基于脑电意念的智能运动耳机实现方法， 其特征在于：耳机控制模块中，耳机控制模块接收识别模块传入的嵌入式指令，并按照如下优先级顺序，向发声单元发出指令，具体为，
    若接收到eSIM模块传入的来电信息，则优先向发声单元发送来电警示信息，并使发声单元暂停当前任务，向用户播放有来电的警示音乐；在向发声单元发送来电警示信息后，耳机控制模块进入等待状态，直至接收到识别模块传入的接听电话的嵌入式指令或挂断电话的嵌入式指令；
    若在等待状态下接收到接听电话的嵌入式指令，则耳机控制模块向eSIM模块发送接听电话的指令，向发声单元发送播放通话语音的信息，随后继续进入等待状态；eSIM模块选择接听电话，而控制器则使耳机开始传输通话语音；
    若在等待状态下接收到挂断电话的嵌入式指令，则耳机控制模块向eSIM模块发送挂断电话的指令，向发声单元发送继续先前任务的指令；
    若在非智能音乐模式下接收到进入智能音乐模式的嵌入式指令，则向发声单元发送播放来自智能音乐控制模块的音乐的指令；若在智能音乐模式下接收到退出智能音乐模式的嵌入式指令，则向发声单元发送播放来自存储模块的音乐的指令；
    若接收到开始音乐、暂停音乐、停止音乐、切换下一首、切换上一首的嵌入式指令，则向发声单元发送相应指令。
  9. 一种采用权利要求4-8任一项所述的基于脑电意念的智能运动耳机的实现方法的基于脑电意念的智能运动耳机的实现系统，其特征在于：包括脑电信号采集模块、脑电信号预处理模块、识别模块、eSIM模块和储存模块，
    脑电信号采集模块：进行脑电信号采集，采用干电极脑电采集技术，由脑电信号采集模块中的四个单数据通道采集的干态电极传感器1a、1b、1c、1d通过设定的采样频率进行脑电信号采集；
    脑电信号预处理模块：对采集的脑电信号进行预处理，由脑电信号预处理模块接收脑电采集模块采集到的脑电信号，脑电信号预处理模块进行脑电信号的滤波与放大；
    识别模块：通过脑电特征提取模型获得分类结果，对步骤S2预处理后的脑电信号，采用结合独立分量分析ICA和小波分析WY的脑电特征提取方法进行特征提取，获得分类结果；建立脑电特征深度学习模型：采用反向传播算法训练前馈神经网络学习模型；对实时采集的脑电信号进行识别，将步骤S3、S4得 到的参数直接预设在识别模块中，将运动者运动过程中实时采集到的脑电信号数据，通过预处理模块预处理后，输入识别模块后，得到相应的分类结果；对耳机控制模块发送指令，经识别模块的处理器运算后得到的分类结果，与存储模块中的分类结果进行比对，得到对应的嵌入式指令，将嵌入式指令发送给耳机控制模块；
    eSIM模块：通过eSIM模块进行通话控制，实现收到来电、接听电话、挂断电话的控制，并将相关控制信息发送给耳机控制模块；
    存储模块：用于存储脑电信号的分类结果以及与其对应的嵌入式指令。
  10. 如权利要求9所述的基于脑电意念的智能运动耳机的实现系统，其特征在于：还包括智能音乐控制模块：接收心率传感器采集的实时心率数据，实现音乐的音量和频率随运动者的实时心率数据的变化而进行相应变化，具体为，
    接收到心率传感器传入的实时心率数据后，对当前心率所代表的运动强度进行实时分析，运动者的运动强度与运动者的实时心率的关系为：
    Figure PCTCN2019092880-appb-100029
    Figure PCTCN2019092880-appb-100030
    式中，R t为有氧运动的实时心率，R m为有氧运动的最大心率，R q为安静心率，Y为运动者的年龄，Q R为由心率测得的运动者的实时运动强度；
    除使用心率衡量运动强度外，还需使用脑电波衡量，使用脑电波相对功率的增长率来衡量运动者的运动强度，其定义如下：
    Figure PCTCN2019092880-appb-100031
    式中，P为运动者安静时的脑电波相对功率，ΔP为运动者运动时脑电波相对功率的变化量；
    使用如下公式最终确定音乐频率和音量的变化情况，实时音乐频率和音量为：
    Figure PCTCN2019092880-appb-100032
    Figure PCTCN2019092880-appb-100033
    式中，f 0表示初始音乐频率，v 0表示初始音乐音量。
PCT/CN2019/092880 2019-03-15 2019-06-26 基于脑电意念的智能运动耳机及其实现方法与系统 WO2020186651A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910201269.9 2019-03-15
CN201910201269.9A CN109743656B (zh) 2019-03-15 2019-03-15 基于脑电意念的智能运动耳机及其实现方法与系统

Publications (1)

Publication Number Publication Date
WO2020186651A1 true WO2020186651A1 (zh) 2020-09-24

Family

ID=66370686

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/092880 WO2020186651A1 (zh) 2019-03-15 2019-06-26 基于脑电意念的智能运动耳机及其实现方法与系统

Country Status (2)

Country Link
CN (1) CN109743656B (zh)
WO (1) WO2020186651A1 (zh)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113002558A (zh) * 2021-03-30 2021-06-22 复旦大学 一种基于脑电信号的残疾人辅助智能驾驶系统和方法
CN113239744A (zh) * 2021-04-25 2021-08-10 西安电子科技大学 基于深度学习的脑电-肌电信号融合识别方法
CN114209340A (zh) * 2021-12-16 2022-03-22 湖南大学 用卷积神经网络识别概念衍生阶段设计思维的系统和方法
CN114431867A (zh) * 2020-11-02 2022-05-06 北京机械设备研究所 一种脑电采集装置、脑电目标识别方法和装置
CN114469134A (zh) * 2021-12-16 2022-05-13 湖南大学 用残差网络识别概念衍生阶段设计思维脑电的系统和方法
CN114489097A (zh) * 2021-12-20 2022-05-13 西安理工大学 基于精密运动手势的无人机飞行姿态脑控方法
CN115154828A (zh) * 2022-08-05 2022-10-11 安徽大学 一种基于脑机接口技术的大脑功能重塑方法、系统、设备
CN116841386A (zh) * 2023-03-06 2023-10-03 湖南工商大学 结合深度学习与运动想象脑机接口的意念观影系统及方法

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109743656B (zh) * 2019-03-15 2023-04-11 南京邮电大学 基于脑电意念的智能运动耳机及其实现方法与系统
CN110163436A (zh) * 2019-05-23 2019-08-23 西北工业大学 基于动态瓶颈预测的智能车间生产优化方法
CN110200625A (zh) * 2019-07-05 2019-09-06 郭长娥 一种大脑脑部神经强化训练装置及方法
CN110324774A (zh) * 2019-07-18 2019-10-11 河海大学常州校区 基于脑电eeg的骨传导助听器自动增益控制系统
CN110650408B (zh) * 2019-10-29 2022-04-19 歌尔股份有限公司 一种耳机及防丢方法、系统、计算机可读存储介质
CN110974219A (zh) * 2019-12-20 2020-04-10 北京脑陆科技有限公司 一种基于侵入式bci的人脑意念识别系统
CN112799505B (zh) * 2020-12-21 2022-08-26 杭州电子科技大学 一种基于脑电信号的音乐音调想象区分方法
CN112969117B (zh) * 2021-02-02 2022-05-17 深圳市宏智力科技有限公司 多功能脑波智能设备和多功能脑波耳机

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6154669A (en) * 1998-11-06 2000-11-28 Capita Systems, Inc. Headset for EEG measurements
CN101515200A (zh) * 2009-04-03 2009-08-26 北京工业大学 基于瞬态视觉诱发脑电的目标选择方法
CN105446492A (zh) * 2015-12-29 2016-03-30 武汉华星光电技术有限公司 基于脑电波感应耳机和智能穿戴设备的信息交互系统
CN208434108U (zh) * 2018-06-22 2019-01-25 罗令 具有eSIM卡功能的发夹助听器
CN109743656A (zh) * 2019-03-15 2019-05-10 南京邮电大学 基于脑电意念的智能运动耳机及其实现方法与系统

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105147282A (zh) * 2015-08-25 2015-12-16 上海医疗器械高等专科学校 认知功能障碍脑电识别系统
CN107527634A (zh) * 2017-07-14 2017-12-29 昆明理工大学 一种脑电波控制音乐播放的装置
CN207099301U (zh) * 2017-07-26 2018-03-13 歌尔科技有限公司 一种无线耳机
CN108834012B (zh) * 2018-03-27 2021-06-22 江苏大学 一种能够提高运动质量的蓝牙耳机系统及其控制方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6154669A (en) * 1998-11-06 2000-11-28 Capita Systems, Inc. Headset for EEG measurements
CN101515200A (zh) * 2009-04-03 2009-08-26 北京工业大学 基于瞬态视觉诱发脑电的目标选择方法
CN105446492A (zh) * 2015-12-29 2016-03-30 武汉华星光电技术有限公司 基于脑电波感应耳机和智能穿戴设备的信息交互系统
CN208434108U (zh) * 2018-06-22 2019-01-25 罗令 具有eSIM卡功能的发夹助听器
CN109743656A (zh) * 2019-03-15 2019-05-10 南京邮电大学 基于脑电意念的智能运动耳机及其实现方法与系统

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114431867A (zh) * 2020-11-02 2022-05-06 北京机械设备研究所 一种脑电采集装置、脑电目标识别方法和装置
CN114431867B (zh) * 2020-11-02 2024-05-10 北京机械设备研究所 一种脑电采集装置、脑电目标识别方法和装置
CN113002558A (zh) * 2021-03-30 2021-06-22 复旦大学 一种基于脑电信号的残疾人辅助智能驾驶系统和方法
CN113239744B (zh) * 2021-04-25 2024-04-19 西安电子科技大学 基于深度学习的脑电-肌电信号融合识别方法
CN113239744A (zh) * 2021-04-25 2021-08-10 西安电子科技大学 基于深度学习的脑电-肌电信号融合识别方法
CN114209340A (zh) * 2021-12-16 2022-03-22 湖南大学 用卷积神经网络识别概念衍生阶段设计思维的系统和方法
CN114469134A (zh) * 2021-12-16 2022-05-13 湖南大学 用残差网络识别概念衍生阶段设计思维脑电的系统和方法
CN114489097A (zh) * 2021-12-20 2022-05-13 西安理工大学 基于精密运动手势的无人机飞行姿态脑控方法
CN114489097B (zh) * 2021-12-20 2023-06-30 西安理工大学 基于精密运动手势的无人机飞行姿态脑控方法
CN115154828A (zh) * 2022-08-05 2022-10-11 安徽大学 一种基于脑机接口技术的大脑功能重塑方法、系统、设备
CN115154828B (zh) * 2022-08-05 2023-06-30 安徽大学 一种基于脑机接口技术的大脑功能重塑方法、系统、设备
CN116841386A (zh) * 2023-03-06 2023-10-03 湖南工商大学 结合深度学习与运动想象脑机接口的意念观影系统及方法
CN116841386B (zh) * 2023-03-06 2024-05-24 湖南工商大学 结合深度学习与运动想象脑机接口的意念观影系统及方法

Also Published As

Publication number Publication date
CN109743656B (zh) 2023-04-11
CN109743656A (zh) 2019-05-10

Similar Documents

Publication Publication Date Title
WO2020186651A1 (zh) 基于脑电意念的智能运动耳机及其实现方法与系统
CN106725462B (zh) 基于脑电信号的声光睡眠干预系统和方法
WO2021103084A1 (zh) 一种用于睡眠调节的深度声音刺激系统和方法
CN110353702A (zh) 一种基于浅层卷积神经网络的情感识别方法及系统
CN107280694A (zh) 一种基于多源信息融合的疲劳检测方法
CN101576772B (zh) 基于虚拟仪器稳态视觉诱发电位脑机接口系统及控制方法
WO2021103829A1 (zh) 基于脑波音乐的个性化精神状态调节系统及调节方法
US20210128049A1 (en) Pronunciation function evaluation system based on array high-density surface electromyography
CN106267514B (zh) 基于脑电反馈的情绪调控系统
CN106345034A (zh) 基于脑电采集终端的用于认知情绪调节的装置
CN110947076B (zh) 一种可进行精神状态调节的智能脑波音乐可穿戴设备
WO2020187109A1 (zh) 一种用户睡眠检测方法及系统
CN110368005A (zh) 一种智能耳机及基于智能耳机的情绪及生理健康监控方法
CA2993636C (en) Devices, systems and methods for monitoring brain activity and for enabling brain to rebalance
CN104571504B (zh) 一种基于想象动作的在线脑‑机接口方法
US20180353128A1 (en) Active unipolar dry electrode open ear wireless headset and brain computer interface
WO2016173374A1 (zh) 基于脑电波无线耳机的音乐终端实时交互系统
CN110074779A (zh) 一种脑电信号识别方法及装置
CN110742603A (zh) 一种脑电波可听化精神状态检测方法及实现该方法的系统
CN104257380A (zh) 一种脑电图采集处理系统
CN107184205B (zh) 基于脑的记忆刻度和诱导捕捉的自动知识记忆牵引方法
CN205924014U (zh) 一种睡眠综合监护系统
Ferracuti et al. Auditory paradigm for a P300 BCI system using spatial hearing
CN109567936A (zh) 一种基于听觉注意和多焦电生理的脑机接口系统和实现方法
CN207666587U (zh) 脑电波眼罩及脑电波信号采集装置

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: 19920039

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19920039

Country of ref document: EP

Kind code of ref document: A1