Multi-channel weak signal acquisition system

Abstract

The present application relates to the technical field of engineering equipment, and specifically discloses a multi-channel weak signal acquisition system, which uses the ADS1299 chip to achieve high-precision signal acquisition, and reduces signal interference by designing hardware circuits and software algorithms. The hardware circuit uses ESD protection circuits and RC filter circuits for signal preprocessing to ensure the stability and anti-interference ability of the signal during transmission. The software part introduces deep learning technology to perform time-frequency analysis, feature filtering, and signal reconstruction on the human physiological signal data transmitted by the hardware circuit, thereby improving the signal quality. In this way, the signal interference problem in microfluidic chip detection can be effectively solved, and the detection accuracy and stability can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of engineering equipment, and more specifically, to a multi-channel weak signal acquisition system. Background Art

[0002] Microfluidics refers to the technology of processing or manipulating fluids in microchannels with a size of tens to hundreds of microns. This technology involves disciplines such as fluid physics, microelectronics, and biomedical engineering. Microfluidics has the advantages of low cost and good performance, so it is widely used in the fields of microelectronics, biochemistry, and medicine.

[0003] However, while microfluidic chip detection requires extremely high detection accuracy, the detected object itself also has a large interference signal, which places high demands on the accuracy and stability of signal detection. The detection accuracy of the traditional multi-channel probe station signal detection system is not enough to cope with high-precision microfluidic chip signal detection. High-precision signal detection instruments can only detect signals from general devices, and it is difficult to achieve signal detection of precision chips, especially microfluidic chips.

[0004] Therefore, a high-precision and high-stability multi-channel weak signal acquisition system is expected. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of the present application provides a multi-channel weak signal acquisition system, which uses the ADS1299 chip to achieve high-precision signal acquisition, and reduces signal interference by designing hardware circuits and software algorithms. Among them, the hardware circuit uses ESD protection circuits and RC filter circuits for signal preprocessing to ensure the stability and anti-interference ability of the signal during transmission. The software part introduces deep learning technology to perform time-frequency analysis, feature filtering and signal reconstruction on the human physiological signal data transmitted by the hardware circuit, thereby improving the signal quality. In this way, the signal interference problem in microfluidic chip detection can be effectively solved, and the detection accuracy and stability can be improved.

[0006] Accordingly, according to one aspect of the present application, a multi-channel weak signal acquisition system is provided, comprising:

[0007] Multi-channel weak signal acquisition hardware circuit, used to simultaneously collect human physiological signal data from multiple channels;

[0008] A real-time data receiving module, used for receiving human physiological signal data of the multiple channels transmitted by the multi-channel weak signal acquisition hardware circuit;

[0009] A data processing module, used for performing time-frequency feature filtering and signal reconstruction on the human physiological signal data of each channel to obtain processed human physiological signal data of multiple channels;

[0010] The data storage module is used to store the processed data of human physiological signals of the multiple channels.

[0011] In the above-mentioned multi-channel weak signal acquisition system, the multi-channel weak signal acquisition hardware circuit includes a pre-filtering circuit, a downloading circuit, a minimum system part for controlling a single-chip microcomputer, an ASD1299 signal detection circuit, a power supply circuit, and a charging circuit; the pre-filtering circuit is used to pre-filter the collected human physiological signal data to obtain the pre-filtered human physiological signal data; the downloading circuit is used to isolate and protect the human physiological signal data to ensure the signal integrity of the human physiological signal data; the ASD1299 signal detection circuit is used to convert the human physiological signal data from an analog signal to a digital signal; the power supply circuit is used to adjust the output voltage; and the charging circuit is used to perform charging management.

[0012] In the above-mentioned multi-channel weak signal acquisition system, the pre-filter circuit includes 8 headphone sockets, and the 8 headphone sockets are used to connect sensor leads to collect the human physiological signal data, wherein each of the headphone sockets is equipped with an ESD diode for electrostatic protection and an RC filtering circuit.

[0013] In the above multi-channel weak signal acquisition system, the minimum system part of the control single-chip microcomputer includes a single-chip microcomputer, a crystal oscillator circuit, a reset circuit and a power supply.

[0014] In the above-mentioned multi-channel weak signal acquisition system, the data processing module includes: a physiological signal time-frequency feature extraction unit, which is used to extract the time-frequency features of the human physiological signal data of each channel to obtain a time-frequency feature graph of the human physiological signal; a feature dispersion unit, which is used to perform feature dispersion on the time-frequency feature graph of the human physiological signal along the channel dimension to obtain a sequence of time-frequency feature vectors of the human physiological signal channel dimension; a feature filtering unit, which is used to perform feature filtering based on semantic contribution on the sequence of time-frequency feature vectors of the human physiological signal channel dimension to obtain a sequence of filtered time-frequency feature vectors of the human physiological signal channel dimension; a signal reconstruction unit, which is used to perform signal restoration processing based on the sequence of filtered time-frequency feature vectors of the human physiological signal channel dimension to obtain processed data of the human physiological signals of each channel.

[0015] In the above-mentioned multi-channel weak signal acquisition system, the physiological signal time-frequency feature extraction unit includes: a wavelet analysis subunit, which is used to perform wavelet analysis on the human physiological signal data of each channel to obtain a human physiological signal time-frequency diagram; a time-frequency feature extraction subunit, which is used to input the human physiological signal time-frequency diagram into a human physiological signal time-frequency feature extractor based on a void convolutional neural network model to obtain the human physiological signal time-frequency feature diagram.

[0016] In the above-mentioned multi-channel weak signal acquisition system, the feature filtering unit includes: a cluster analysis subunit, which is used to perform cluster analysis on the sequence of time-frequency feature vectors of the human physiological signal channel dimension to obtain a human physiological signal time-frequency feature cluster center representation vector as a pseudo anchor center; a semantic contribution calculation subunit, which is used to calculate the semantic contribution representation matrix of each human physiological signal channel dimension time-frequency feature vector in the sequence of human physiological signal channel dimension time-frequency feature vectors relative to the pseudo anchor center to obtain a sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices; a feature modulation subunit, which is used to perform feature modulation on the sequence of human physiological signal channel dimension time-frequency feature vectors based on the sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices to obtain the sequence of filtered human physiological signal channel dimension time-frequency feature vectors.

[0017] In the above-mentioned multi-channel weak signal acquisition system, the semantic contribution calculation subunit is used to: multiply the time-frequency feature vector of the human physiological signal channel dimension by the transposed vector of the pseudo anchor center and then divide it by the second norm of the pseudo anchor center to obtain the semantic contribution representation matrix of the time-frequency feature of the human physiological signal channel dimension.

[0018] In the above-mentioned multi-channel weak signal acquisition system, the feature modulation subunit is used to: perform contribution mask sparse processing based on the random inactivation mechanism on each human physiological signal channel dimension time-frequency feature semantic contribution representation matrix in the sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices to obtain a sequence of masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrices; use each masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrix in the sequence of masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrices as a modulation matrix, and calculate the matrix product between it and each human physiological signal channel dimension time-frequency feature vector to obtain the sequence of filtered human physiological signal channel dimension time-frequency feature vectors.

[0019] In the above-mentioned multi-channel weak signal acquisition system, the signal reconstruction unit is used to: cascade the sequence of the filtered human physiological signal channel dimension time-frequency feature vectors into a filtered full time-frequency domain human physiological signal aggregate representation vector, and then input the filtered full time-frequency domain human physiological signal aggregate representation vector into a decoder-based signal restorer to obtain the processed human physiological signal data of each channel.

[0020] Compared with the prior art, the multi-channel weak signal acquisition system provided by this application uses the ADS1299 chip to achieve high-precision signal acquisition, and reduces signal interference by designing hardware circuits and software algorithms. The hardware circuit uses ESD protection circuits and RC filter circuits for signal preprocessing to ensure the stability and anti-interference ability of the signal during transmission. The software part introduces deep learning technology to perform time-frequency analysis, feature filtering, and signal reconstruction on the human physiological signal data transmitted by the hardware circuit, thereby improving signal quality. In this way, the signal interference problem in microfluidic chip detection can be effectively solved, and the detection accuracy and stability can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 4 is a block diagram of a multi-channel weak signal acquisition system according to an embodiment of the present application.

[0023] Figure 2 4 is a circuit diagram of a multi-channel weak signal acquisition system according to an embodiment of the present application.

[0024] Figure 3 The present invention is a flowchart of a single-chip microcomputer embedded program of a multi-channel weak signal acquisition system according to an embodiment of the present application.

[0025] Figure 4 4 is a block diagram of a data processing module in a multi-channel weak signal acquisition system according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0027] As mentioned in the above background technology, human physiological electrical signals, as extremely weak signals, are often easily encountered with multiple interferences during the measurement process, thereby affecting the quality and accuracy of the signal. For example, in the process of collecting ECG, the most common interferences mainly include baseline drift, electromagnetic interference, myoelectric interference, motion artifacts and electrostatic interference. In order to overcome the problem of signal interference, in view of the limitations of the prior art, this application proposes a novel multi-channel weak signal acquisition system, which uses ADS1299 chip to achieve high-precision signal acquisition, and reduces signal interference by designing hardware circuits and software algorithms, wherein the hardware circuit uses ESD protection circuits and RC filter circuits for signal preprocessing to ensure the stability and anti-interference ability of the signal during transmission, and the software part introduces deep learning technology to perform time-frequency analysis, feature filtering and signal reconstruction on the human physiological signal data transmitted by the hardware circuit, thereby improving signal quality. In this way, the signal interference problem in microfluidic chip detection can be effectively solved, and detection accuracy and stability can be improved.

[0028] Figure 1 FIG. 1 is a block diagram of a multi-channel weak signal acquisition system according to an embodiment of the present application. Figure 1 As shown, according to the multi-channel weak signal acquisition system 100 of the embodiment of the present application, it includes: a multi-channel weak signal acquisition hardware circuit 110, which is used to simultaneously acquire human physiological signal data of multiple channels; a real-time data receiving module 120, which is used to receive the human physiological signal data of the multiple channels transmitted by the multi-channel weak signal acquisition hardware circuit; a data processing module 130, which is used to perform time-frequency feature filtering and signal reconstruction on the human physiological signal data of each channel to obtain processed human physiological signal data of multiple channels; a data storage module 140, which is used to store the processed human physiological signal data of the multiple channels.

[0029] Specifically, in the design process of the multi-channel weak signal acquisition hardware circuit 110, considering the signal peak value and acquisition sensitivity of the acquisition signal, the present application selected the ADS1299 integrated chip of TI. The ADS1299 is an 8-channel low-noise simultaneous sampling 24-bit delta-sigma ADC, and has a built-in programmable gain amplifier (PGA), reference voltage and oscillator, integrating the general characteristics required for electroencephalogram (EEG). The device can adjust the measurement range according to the VREF input, up to 6.1VPP, programmable gain is 1, 2, 4, 6, 8, 12 or 24, the minimum range is VREF divided by the maximum programmable gain, the maximum sampling rate of a single channel is 16KSPS, and it can use a very low input reference noise of 1.0uVPP (70Hz BW); equipped with a serial peripheral interface (Serial Peripheral Interface, SPI), to achieve full-duplex bidirectional transmission; built-in drive amplifier circuit and common mode rejection circuit, common mode rejection ratio (Common Mode Rejection Ratio, CMRR) greater than 110dB; supports bipolar or unipolar input, power consumption as low as 5mW. It is mainly used in medical instruments such as EEG and ECG, auditory evoked potential (EAP) EEG bispectral index (BIS), sleep research monitoring and other high-precision multi-channel signal acquisition.

[0030] The single-chip microcomputer of the present application is preferably the STM32F103 single-chip microcomputer, which adopts the high-performance, low-cost, low-power ARMCortex-M core, and has first-class peripherals, such as 1μs dual 12-bit ADC, 4 Mbps UART, 18 Mbps SPI, etc. It also has excellent performance in power consumption and integration, and is suitable for applications in a variety of complex scenarios. In this system, the main controller STM32F103ZET6 communicates with slave devices such as ADS1299 through SPI.

[0031] Therefore, the signal acquisition chip of this system is preferably the ADS1299 integrated chip, with a measurement range of up to ±6.1VPP, a resolution that can be adjusted according to the reference voltage input and programmable gain, an adjustable sampling rate from 250Hz to 16kHz, and an input reference noise of 1.0uVPP (70Hz BW). It has the characteristics of high resolution, high precision, high sampling rate, and low noise, and is suitable for accurate measurement of voltage signals in a variety of complex scenarios. The single-chip microcomputer of this system is preferably the STM32F103 single-chip microcomputer, which uses a high-performance, low-power ARM Cortex-M core, has first-class peripherals, and is suitable for responses in a variety of complex scenarios.

[0032] Figure 2 FIG. 1 is a circuit schematic diagram of a multi-channel weak signal acquisition system according to an embodiment of the present application. Figure 2As shown in the figure, the multi-channel weak signal acquisition hardware circuit mainly includes 6 parts, namely, the pre-filter circuit, the download circuit, the minimum system part of the control microcontroller (here, the minimum system of the STM32F103C8 microcontroller is taken as an example), the ASD1299 signal detection circuit, the power supply circuit, and the charging circuit. Among them, the minimum system of the control microcontroller includes the microcontroller, the reset circuit, the crystal oscillator (clock) circuit, and the power supply.

[0033] Specifically, the pre-filter circuit is used to pre-filter the collected human physiological signal data to obtain the pre-filtered human physiological signal data, and it includes 8 headphone sockets, the 8 headphone sockets are used to connect sensor leads to collect the human physiological signal data, and each of the headphone sockets is equipped with an ESD diode for electrostatic protection and an RC filtering circuit.

[0034] The electrophysiological signals of the human body, such as electrocardiogram and electromyography, are very weak, usually in the range of microvolts to millivolts. Therefore, the circuits used to collect these signals must be extremely sensitive and precise, which makes the circuits very sensitive to external interference, especially electrostatic discharge (ESD). Based on this, the present application designs an ESD protection circuit, which uses a transient voltage suppression diode (TVS diode) to protect sensitive electronic equipment from damage caused by electrostatic discharge (ESD) and other transient overvoltage events. Based on the nonlinear voltage-current characteristics of the TVS diode, under normal operating voltage, the TVS diode is in a high-resistance state, and the impact on the circuit is minimal. When the voltage exceeds its predetermined breakdown voltage, the TVS diode quickly turns on and becomes a low-resistance state, thereby quickly directing the excessive voltage to the ground line to protect downstream sensitive components. It can respond to overvoltage events in a very short time and direct the excessive voltage to the ground line, thereby preventing the voltage from penetrating into sensitive components.

[0035] In a specific example of the present application, the RC passive low-pass filter circuit of the pre-filter circuit uses a 22.1 kΩ resistor and a 47 pF capacitor, and the cut-off frequency of the filter circuit is approximately 153 kHz.

[0036] Specifically, the download circuit is used to isolate and protect the human physiological signal data to ensure the signal integrity of the human physiological signal data, and it mainly includes a CH340 USB to serial port chip and an ADUM1201 digital isolator. Among them, the CH340 USB to serial port chip is responsible for transmitting the collected signal data to a computer or other device through a USB interface, while the ADUM1201 digital isolator ensures electrical isolation during signal transmission, avoiding ground loop interference and common mode interference. It should be understood that the purpose of this system is to collect weak electrophysiological signals such as human electrophysiological signals. The digital isolator can ensure the integrity of the signal on the isolation interface to a certain extent, and avoid high-frequency noise and electromagnetic interference affecting the accuracy of the measurement. At the same time, it can isolate data paths and control signals, reduce interference between systems, and improve stability and reliability.

[0037] Specifically, the ASD1299 signal detection circuit is used to convert the human physiological signal data from an analog signal to a digital signal. As a high-performance, low-power multi-channel biological signal acquisition chip, ADS1299 is specifically used for biomedical applications and physiological monitoring. First, the internal design of the preamplifier of ADS1299 adopts a low-noise operational amplifier, which can amplify weak biological signals with low distortion and low noise. Secondly, the 24-bit analog-to-digital converter integrated in the chip can provide high-resolution analog-to-digital conversion, so that even tiny biological signals can be accurately digitized, thereby maintaining the accuracy of the signal. In addition, the chip is equipped with a variety of filters to effectively remove interference from the external environment, and further filter and process the signal in the digital domain, which helps to improve the signal-to-noise ratio. In addition, even if there is electromagnetic interference in the external environment, since ADS1299 adopts a series of anti-interference design measures including shielding and suppression technology, it can greatly reduce the impact of external electromagnetic interference on signal quality when measuring signals.

[0038] Specifically, the power supply circuit is used to adjust the output voltage, which includes the hx4002 boost chip to adjust the output voltage, the TPS73633 low-dropout linear regulator to provide a stable 3.3V power output, and supplies the 3.3V voltage to the VDD pin of the ADS1299, the VDD pin of the microcontroller and other modules. The hx4002 boost chip is connected to the power supply through the VIN pin to provide the DC input voltage required for the chip to work, and outputs the boosted 5V voltage through the VOUT pin, and reduces the high-frequency noise through the parallel bypass capacitor to play a noise reduction filtering role.

[0039] Specifically, the charging circuit is used for charging management, which mainly includes a tp4056 charging management chip and supporting circuits, and an external charging and discharging lithium battery is connected to the HC-1.25-2A terminal (U10 interface) of the power module as an external power supply for the system.

[0040] Figure 3 FIG. 1 is a flow chart of a single-chip embedded program of a multi-channel weak signal acquisition system according to an embodiment of the present application. Figure 3 As shown, first, when the circuit board is powered on, the control microcontroller and the ADS1299 integrated chip on the circuit board automatically enter the initialization program; the control microcontroller initializes the USART serial communication, establishes the serial communication connection with the host computer, and the control microcontroller initializes the SPI communication, and establishes the communication based on the SPI communication line with the ADS1299 capacitance detection chip; after completing the communication initialization, the microcontroller initializes the timer and enables the timer count. The microcontroller sends a read register instruction to the ADS1299 integrated chip to confirm the chip operation status, and then sends a write register instruction to set the chip status. By setting the ADS1299 register, the parameters such as the ADS1299 selection channel, channel detection mode, and pin setting are initialized. Then, the timer count is read and cleared. The chip is in the discontinuous data reading mode at this time. The microcontroller sends a read data command to the ADS1299 integrated chip and receives the data packet sent by the ADS1299 on the data communication line. Each data packet contains the signal data of eight channels. The four-byte data of each channel of the eight channels are integrated and recalculated as floating-point data. The channel number is added to the data. The signal detection time is calculated according to the timer data. The ADS1299 data packet is disassembled to obtain the specific data of the signal values ​​of different channels. The loop count value, channel flag, and channel signal data are repackaged and sent to the host computer through the USART serial port communication. This process is repeated until the program terminates.

[0041] After the real-time data receiving module 120 deployed on the host computer receives the human physiological signal data of the multiple channels transmitted by the multi-channel weak signal acquisition hardware circuit, in order to effectively extract useful signals from the data and suppress interference, the present application further adopts a signal time-frequency analysis method based on deep learning technology to process the collected human physiological signal data.

[0042] Figure 4 FIG. 1 is a block diagram of a data processing module in a multi-channel weak signal acquisition system according to an embodiment of the present application. Figure 4As shown, the data processing module 130 includes: a physiological signal time-frequency feature extraction unit 131, which is used to extract the time-frequency features of the human physiological signal data of each channel to obtain a human physiological signal time-frequency feature diagram; a feature dispersion unit 132, which is used to perform feature dispersion on the human physiological signal time-frequency feature diagram along the channel dimension to obtain a sequence of human physiological signal channel dimension time-frequency feature vectors; a feature filtering unit 133, which is used to perform feature filtering on the sequence of human physiological signal channel dimension time-frequency feature vectors based on semantic contribution to obtain a sequence of filtered human physiological signal channel dimension time-frequency feature vectors; a signal reconstruction unit 134, which is used to perform signal restoration processing based on the sequence of filtered human physiological signal channel dimension time-frequency feature vectors to obtain the processed data of the human physiological signals of each channel.

[0043] Specifically, the physiological signal time-frequency feature extraction unit 131 is used to extract the time-frequency features of the human physiological signal data of each channel to obtain a time-frequency feature graph of the human physiological signal. In a specific example of the present application, the physiological signal time-frequency feature extraction unit 131 includes: a wavelet analysis subunit, which is used to perform wavelet analysis on the human physiological signal data of each channel to obtain a time-frequency graph of the human physiological signal; and a time-frequency feature extraction subunit, which is used to input the time-frequency graph of the human physiological signal into a human physiological signal time-frequency feature extractor based on a hollow convolutional neural network model to obtain the time-frequency feature graph of the human physiological signal.

[0044] It should be understood that human physiological signals (such as electrocardiogram (ECG), electroencephalogram (EEG), etc.) are usually non-stationary signals, and their frequency components change over time. As an effective time-frequency analysis method, wavelet analysis can effectively capture the frequency distribution of signals at different time scales. Compared with traditional methods such as Fourier transform, wavelet analysis has obvious advantages in processing non-stationary signals, and can provide detailed information of signals in both time domain and frequency domain, thus providing important basic data for subsequent time-frequency feature extraction and signal filtering.

[0045] Next, in order to further extract representative time-frequency features from the time-frequency diagram of the human physiological signal obtained by wavelet analysis, the present application adopts a dilated convolutional neural network model with excellent performance in the field of image processing as a time-frequency feature extractor of a human physiological signal to process the time-frequency diagram of the human physiological signal. The dilated convolution characteristics are utilized to expand the receptive field while maintaining a high spatial resolution, so as to extract local details and broader contextual information in the time-frequency diagram of the human physiological signal, thereby capturing rapidly changing frequency components (such as spikes) and slowly changing trends (such as baseline drift) therein, so as to generate a time-frequency feature diagram of the human physiological signal.

[0046] Specifically, the feature dispersion unit 132 is used to perform feature dispersion on the time-frequency feature graph of the human physiological signal along the channel dimension to obtain a sequence of time-frequency feature vectors of the human physiological signal channel dimension. It should be understood that in order to filter out interference components in the signal and improve signal quality, it is necessary to further perform feature filtering on the time-frequency feature graph of the human physiological signal. In the technical solution of the present application, firstly, by performing feature dispersion processing on the time-frequency feature graph of the human physiological signal along the channel dimension, the feature graph is decomposed into a series of local time-frequency feature vector forms of the channel dimension, so as to further refine the accuracy of feature filtering, and then by performing feature importance analysis on the time-frequency feature vectors of the channel dimension of each human physiological signal, the time-frequency features of each channel dimension are screened and suppressed, thereby filtering out interference components and retaining useful features.

[0047] Specifically, the feature filtering unit 133 is used to perform feature filtering based on semantic contribution on the sequence of time-frequency feature vectors of the human physiological signal channel dimension to obtain a sequence of time-frequency feature vectors of the filtered human physiological signal channel dimension. That is, in the technical solution of the present application, the sequence of time-frequency feature vectors of the human physiological signal channel dimension is subjected to cluster analysis and semantic contribution evaluation to achieve feature screening and optimization. Specifically, first, the overall internal structure and distribution of the data are captured by clustering analysis on the sequence of time-frequency feature vectors of the human physiological signal channel dimension, and the human physiological signal time-frequency feature cluster center representation vector is extracted as a pseudo anchor center, thereby providing a reference for subsequent feature evaluation. Then, by calculating the semantic contribution of each human physiological signal channel dimension time-frequency feature vector relative to the pseudo anchor center, the correlation and relative importance of each human physiological signal channel dimension time-frequency feature vector and the trend of the data set are quantified to generate a sequence of semantic contribution representation matrices. Then, by performing random inactivation mask processing on each semantic contribution representation matrix, the focus on key features is strengthened, the generalization ability of the model is enhanced, and the number of parameters is reduced. Finally, the semantic contribution representation matrices after masking are used as modulation matrices, and the original time-frequency feature vectors of each human physiological signal channel dimension are weighted to filter out unimportant or noise features, thereby obtaining a sequence of filtered time-frequency feature vectors of the human physiological signal channel dimension.

[0048] In a specific example of the present application, the feature filtering unit 133 includes: a cluster analysis subunit, which is used to perform cluster analysis on the sequence of time-frequency feature vectors of the human physiological signal channel dimension to obtain a human physiological signal time-frequency feature cluster center representation vector as a pseudo anchor center; a semantic contribution calculation subunit, which is used to calculate the semantic contribution representation matrix of each human physiological signal channel dimension time-frequency feature vector in the sequence of time-frequency feature vectors of the human physiological signal channel dimension relative to the pseudo anchor center to obtain a sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices; a feature modulation subunit, which is used to perform feature modulation on the sequence of time-frequency feature vectors of the human physiological signal channel dimension based on the sequence of semantic contribution representation matrices of the human physiological signal channel dimension time-frequency feature to obtain the sequence of filtered human physiological signal channel dimension time-frequency feature vectors.

[0049] More specifically, the cluster analysis subunit is used to calculate the mean vector of the sequence of time-frequency feature vectors of the human physiological signal channel dimension as the center representation vector of the clustering of the time-frequency feature of the human physiological signal.

[0050] More specifically, the semantic contribution calculation subunit is used to: multiply the time-frequency feature vector of the human physiological signal channel dimension by the transposed vector of the pseudo anchor center and then divide it by the binary norm of the pseudo anchor center to obtain the semantic contribution representation matrix of the time-frequency feature of the human physiological signal channel dimension.

[0051] More specifically, the feature modulation subunit is used to: perform contribution masked sparsification processing based on a random inactivation mechanism on each human physiological signal channel dimension time-frequency feature semantic contribution representation matrix in the sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices to obtain a sequence of masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrices; use each masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrix in the sequence of masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrices as a modulation matrix, and calculate the matrix product between it and each human physiological signal channel dimension time-frequency feature vector to obtain the sequence of filtered human physiological signal channel dimension time-frequency feature vectors.

[0052] That is, the sequence of the time-frequency feature vectors of the human physiological signal channel dimension is processed by the following feature filtering formula to obtain the sequence of the filtered time-frequency feature vectors of the human physiological signal channel dimension, wherein the feature filtering formula is:

[0053]

[0054]

[0055]

[0056]

[0057] in, A sequence representing the time-frequency feature vector of the human physiological signal channel dimension, , , and are respectively the first, second, and third in the sequence of the time-frequency feature vector of the human physiological signal channel dimension. and Individual physiological signal channel dimension time-frequency feature vector, The value of is the number of time-frequency feature vectors of the human physiological signal channel dimension, represents the pseudo anchor center, represents the bi-norm of the eigenvector, represents the transpose of a vector, is the normalized exponential function, represents random dropout processing, Indicates The semantic contribution representation matrix of the time-frequency feature of the channel dimension of the masked sparse human physiological signal, represents matrix multiplication, Indicates The filtered human physiological signal channel dimension time-frequency feature vector.

[0058] Specifically, the signal reconstruction unit 134 is used to perform signal restoration processing based on the sequence of time-frequency feature vectors of the filtered human physiological signal channel dimension to obtain the processed data of the human physiological signal of each channel. In a specific example of the present application, the processing method for signal restoration based on the sequence of time-frequency feature vectors of the filtered human physiological signal channel dimension is to cascade the sequence of time-frequency feature vectors of the filtered human physiological signal channel dimension into a filtered full-time-frequency domain human physiological signal aggregate representation vector, and then input the filtered full-time-frequency domain human physiological signal aggregate representation vector into a decoder-based signal restorer to obtain the processed data of the human physiological signal of each channel. That is, by cascading the sequence of time-frequency feature vectors of the filtered human physiological signal channel dimension to integrate the signal time-frequency information of each channel. Then, a signal restorer is constructed using a deep learning-based decoder structure, and the filtered full-time-frequency domain human physiological signal aggregate representation vector obtained by cascading is input into the decoder-based signal restorer. The powerful decoding and regression capability of the decoder model is utilized to reconstruct the signal by learning the signal time-frequency features contained in the vector, and the signal time-frequency encoding features are decoded back to the original signal form to restore the waveform of the original signal, thereby obtaining the processed data of the human physiological signals of each channel.

[0059] Preferably, the filtered full time-frequency domain human physiological signal aggregation representation vector is input into a decoder-based signal restorer to obtain the processed data of the human physiological signals of each channel, including: calculating the absolute value sum of each eigenvalue of the filtered full time-frequency domain human physiological signal aggregation representation vector to obtain the first filtered full time-frequency domain human physiological signal aggregation and modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the filtered full time-frequency domain human physiological signal aggregation representation vector to obtain the second filtered full time-frequency domain human physiological signal aggregation and modulation value; subtracting the filtered full time-frequency domain human physiological signal aggregation representation vector from the second filtered full time-frequency domain human physiological signal aggregation and modulation value, and then performing dot multiplication with the number of eigenvalues ​​of the filtered full time-frequency domain human physiological signal aggregation representation vector and the inverse of the first filtered full time-frequency domain human physiological signal aggregation and modulation value, and taking the inverse of each eigenvalue to obtain the first filtered full time-frequency domain human physiological signal aggregation and modulation value. Signal aggregation phase conversion vector; after point subtraction of the filtered full time-frequency domain human physiological signal aggregation representation vector and the first filtered full time-frequency domain human physiological signal aggregation and modulation value, respectively perform point multiplication with the square root of the number of eigenvalues ​​of the filtered full time-frequency domain human physiological signal aggregation representation vector and the inverse of the second filtered full time-frequency domain human physiological signal aggregation and modulation value, and take the inverse of each eigenvalue to obtain the second filtered full time-frequency domain human physiological signal aggregation phase conversion vector; calculate the point product vector of the second filtered full time-frequency domain human physiological signal aggregation phase conversion vector and the weighted hyperparameter, subtract the point product vector from the first filtered full time-frequency domain human physiological signal aggregation phase conversion vector to obtain the optimized filtered full time-frequency domain human physiological signal aggregation representation vector; input the optimized filtered full time-frequency domain human physiological signal aggregation representation vector into a decoder-based signal restorer to obtain the processed human physiological signal data of each channel.

[0060] That is, the filtered full-time-frequency domain human physiological signal aggregate representation vector is processed by the following optimization formula to obtain an optimized filtered full-time-frequency domain human physiological signal aggregate representation vector, wherein the optimization formula is:

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067] in, represents the aggregate representation vector of the filtered full-time and frequency domain human physiological signal, The first one represents the aggregate representation vector of the filtered full-time-frequency domain human physiological signal eigenvalues, The value of is the number of eigenvalues ​​in the aggregate representation vector of the human physiological signal in the full time-frequency domain after filtering, represents the absolute value symbol, is the aggregation and modulation value of the full time-frequency domain human physiological signal after the first filtering, is the second filtered full-time-frequency domain human physiological signal aggregation and modulation value, represents dot product, Indicates point reduction, It means to calculate the inverse of each eigenvalue in the vector. represents the aggregated phase transformation vector of the human physiological signal in the full time-frequency domain after the first filtering, represents the aggregated phase transformation vector of the full-time-frequency domain human physiological signal after the second filtering, is the weighted hyperparameter, The optimized filtered full-time-frequency domain human physiological signal aggregation representation vector.

[0068] That is, considering that the sequence of filtered human physiological signal channel dimension time-frequency feature vectors represents the local time-frequency features of the human physiological signal data of each channel, the filtered time-frequency features are based on the pseudo-anchor center contribution semantic query, when the sequence of filtered human physiological signal channel dimension time-frequency feature vectors is cascaded into a filtered full time-frequency domain human physiological signal aggregation representation vector, the semantic filtering query difference based on the local time-frequency features will make the filtered full time-frequency domain human physiological signal aggregation representation vector as a feature distribution as a whole have full time-frequency domain aggregation differences. Therefore, it is expected to improve the detailed time-frequency semantic aggregation expression effect of the filtered full time-frequency domain human physiological signal aggregation representation vector based on the feature cascade aggregation differences of the local time-frequency domain in the full time-frequency domain.

[0069] Based on this, in the preferred embodiment, the difference of the characteristic value of the filtered full time-frequency domain human physiological signal aggregation representation vector relative to the feature set of the vector as a whole of the filtered full time-frequency domain human physiological signal aggregation representation vector and the modulation representation is used as the semantic change intensity information, and a phase-like transformation corresponding to the position-based intensity modulation is performed through different modulation representation forms, so that the aggregation enhancement of semantic change phase perception can enhance the axial aggregation receptive field along the feature aggregation direction by performing a spatial translation operation based on alternating stacking under the scale equalization of the vector set of the filtered full time-frequency domain human physiological signal aggregation representation vector, thereby enhancing the perception effect of the aggregation semantics of the filtered full time-frequency domain human physiological signal aggregation representation vector on the detailed time-frequency semantic changes, so as to improve the expression effect of the filtered full time-frequency domain human physiological signal aggregation representation vector, and improve the accuracy and signal quality of the processed data of the human physiological signals of each channel obtained by the decoder-based signal restorer input.

[0070] After the signal processing is completed, the processed data of the human physiological signals of the multiple channels are transmitted to the data storage module 140 for storage, so as to facilitate subsequent signal analysis and diagnosis. In the embodiment of the present application, in order to ensure the integrity and accuracy of the data, the present application also adopts a redundant storage mechanism to back up the processed data of the human physiological signals of the multiple channels on multiple storage media.

[0071] In summary, the multi-channel weak signal acquisition system according to the embodiment of the present application is explained, which uses the ADS1299 chip to achieve high-precision signal acquisition, and reduces signal interference by designing hardware circuits and software algorithms. Among them, the hardware circuit uses ESD protection circuits and RC filter circuits for signal preprocessing to ensure the stability and anti-interference ability of the signal during transmission. The software part introduces deep learning technology to perform time-frequency analysis, feature filtering and signal reconstruction on the human physiological signal data transmitted by the hardware circuit, thereby improving the signal quality. In this way, the signal interference problem in microfluidic chip detection can be effectively solved, and the detection accuracy and stability can be improved.

[0072] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0073] In the above embodiments, the description of each embodiment has its own emphasis. For the parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0074] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0075] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0076] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A multi-channel weak signal acquisition system, characterized in that: include: Multi-channel weak signal acquisition hardware circuit, used to simultaneously collect human physiological signal data from multiple channels; A real-time data receiving module, used for receiving human physiological signal data of the multiple channels transmitted by the multi-channel weak signal acquisition hardware circuit; A data processing module, used for performing time-frequency feature filtering and signal reconstruction on the human physiological signal data of each channel to obtain processed human physiological signal data of multiple channels; A data storage module, used to store the processed data of human physiological signals of the multiple channels; Wherein, the data processing module includes: A physiological signal time-frequency feature extraction unit, used to extract the time-frequency features of the human physiological signal data of each channel to obtain a human physiological signal time-frequency feature graph; A feature dispersion unit, used for performing feature dispersion on the time-frequency feature graph of the human physiological signal along the channel dimension to obtain a sequence of time-frequency feature vectors of the human physiological signal channel dimension; A feature filtering unit, used for performing feature filtering based on semantic contribution on the sequence of time-frequency feature vectors of the human physiological signal channel dimension to obtain a sequence of filtered time-frequency feature vectors of the human physiological signal channel dimension; A signal reconstruction unit, used for performing signal restoration processing based on the sequence of time-frequency feature vectors of the filtered human physiological signal channel dimension to obtain processed data of the human physiological signal of each channel; Wherein, the feature filtering unit comprises: A cluster analysis subunit, used for performing cluster analysis on the sequence of the time-frequency feature vectors of the human physiological signal channel dimension to obtain a cluster center representation vector of the time-frequency feature of the human physiological signal as a pseudo anchor center; A semantic contribution calculation subunit, used to calculate the semantic contribution representation matrix of each human physiological signal channel dimension time-frequency feature vector in the sequence of human physiological signal channel dimension time-frequency feature vectors relative to the pseudo anchor center to obtain a sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices; A feature modulation subunit, configured to perform feature modulation on the sequence of time-frequency feature vectors in the dimension of human physiological signal channel based on the sequence of semantic contribution representation matrices of the dimension of time-frequency feature of human physiological signal channel to obtain the sequence of time-frequency feature vectors in the dimension of filtered human physiological signal channel; Wherein, the characteristic modulation subunit is used for: Performing contribution mask sparse processing based on a random dropout mechanism on each human physiological signal channel dimension time-frequency feature semantic contribution representation matrix in the sequence of human physiological signal channel dimension time-frequency feature semantic contribution representation matrices to obtain a sequence of mask sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrices; Each masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrix in the sequence of the masked sparse human physiological signal channel dimension time-frequency feature semantic contribution representation matrices is used as a modulation matrix, and the matrix product between it and each human physiological signal channel dimension time-frequency feature vector is calculated respectively to obtain the sequence of filtered human physiological signal channel dimension time-frequency feature vectors.

2. The multi-channel weak signal acquisition system according to claim 1, characterized in that: The multi-channel weak signal acquisition hardware circuit includes a pre-filter circuit, a download circuit, a minimum system part of a control single-chip microcomputer, an ASD1299 signal detection circuit, a power supply circuit, and a charging circuit; The pre-filter circuit is used to pre-filter the collected human physiological signal data to obtain pre-filtered human physiological signal data; The download circuit is used to isolate and protect the human physiological signal data to ensure the signal integrity of the human physiological signal data; The ASD1299 signal detection circuit is used to convert the human physiological signal data from an analog signal to a digital signal; The power supply circuit is used to adjust the output voltage; The charging circuit is used for charging management.

3. The multi-channel weak signal acquisition system according to claim 2, characterized in that: The pre-filter circuit includes 8 headphone sockets, and the 8 headphone sockets are used to connect sensor leads to collect the human physiological signal data, wherein each of the headphone sockets is equipped with an ESD diode for electrostatic protection and an RC filtering circuit.

4. The multi-channel weak signal acquisition system according to claim 3, characterized in that: The minimum system part of the control single chip microcomputer includes a single chip microcomputer, a crystal oscillator circuit, a reset circuit and a power supply.

5. The multi-channel weak signal acquisition system according to claim 1, characterized in that: The physiological signal time-frequency feature extraction unit comprises: A wavelet analysis subunit, used for performing wavelet analysis on the human physiological signal data of each channel to obtain a time-frequency diagram of the human physiological signal; The time-frequency feature extraction subunit is used to input the time-frequency diagram of the human physiological signal into a time-frequency feature extractor of the human physiological signal based on the hole convolutional neural network model to obtain the time-frequency feature diagram of the human physiological signal.

6. The multi-channel weak signal acquisition system according to claim 5, characterized in that: The semantic contribution calculation subunit is used to: The time-frequency feature vector of the human physiological signal channel dimension is multiplied by the transposed vector of the pseudo anchor center and then divided by the second norm of the pseudo anchor center to obtain the semantic contribution representation matrix of the time-frequency feature of the human physiological signal channel dimension.

7. The multi-channel weak signal acquisition system according to claim 6, characterized in that: The signal reconstruction unit is used for: After the sequence of the filtered human physiological signal channel dimension time-frequency feature vectors is cascaded into a filtered full time-frequency domain human physiological signal aggregate representation vector, the filtered full time-frequency domain human physiological signal aggregate representation vector is input into a decoder-based signal restorer to obtain the processed data of the human physiological signals of each channel.

Images