Adaptive compression method based on difference filtering neural signal action potential detection and related equipment

Abstract

The embodiment of the application provides a kind of based on difference filtering neural signal action potential detection adaptive compression method and related equipment, belong to biological signal detection technical field.The method includes obtaining high flux neural signal, wherein, high flux neural signal includes at least one potential signal, potential signal includes action potential and local field potential;Based on the difference operation detection of high flux neural signal to high frequency narrow pulse mutation characteristics, obtain the action position information of action potential in high flux neural signal;According to action position information, high flux neural signal is carried out event-driven adaptive variable rate compression, and compressed data is obtained.The embodiment of the application can realize the accurate detection of action potential, while greatly reducing resource occupation.

Description

Technical Field

[0001] This application relates to the field of biosignal detection technology, and in particular to an adaptive compression method and related equipment for detecting action potentials of neural signals based on differential filtering. Background Technology

[0002] As the most complex and still not fully understood organ in living organisms, the brain has always been a crucial area of ​​cutting-edge scientific research, with the study of its structure and function being a key focus. Brain-computer interfaces (BCIs), serving as a bridge connecting the brain to external devices, have become a core technology in interdisciplinary research across neuroscience, medical engineering, and artificial intelligence. Building on this foundation, research into BCIs is gradually expanding from primates to marine organisms.

[0003] However, due to limitations in wireless communication bandwidth, high-pass filtering of multi-channel neural data consumes excessive hardware resources, especially when processing thousands of channels in parallel, which makes the system unable to handle it. Therefore, real-time transmission of high-channel-count neural signals requires a large amount of resources, resulting in a bulky implantable device that severely compresses the limited cranial space of marine organisms and affects their natural behavior. At the same time, the extremely high power consumption will rapidly consume the limited battery energy, and frequent charging and maintenance are not possible in the marine environment. This will result in a short battery life, making it difficult to meet the long-term needs of marine life monitoring. Summary of the Invention

[0004] The main objective of this application is to propose an adaptive compression method and related equipment for action potential detection of neural signals based on differential filtering, aiming to achieve accurate detection of action potentials while significantly reducing resource consumption.

[0005] To achieve the above objectives, one aspect of this application proposes an adaptive compression method based on differential filtering of neural signal action potential detection, the method comprising:

[0006] Acquiring high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, the potential signal including action potential and local field potential;

[0007] The high-throughput neural signal is detected by difference operation based on the high-frequency narrow pulse mutation characteristics to obtain the action position information of the action potential in the high-throughput neural signal;

[0008] Based on the action location information, the high-throughput neural signal is subjected to event-driven adaptive variable rate compression to obtain compressed data.

[0009] In some embodiments, the step of performing difference calculations on the high-throughput neural signal based on the high-frequency narrow pulse mutation characteristics to obtain the action potential's action location information in the high-throughput neural signal includes the following steps:

[0010] The high-throughput neural signal is subjected to differential processing to obtain difference data;

[0011] Dynamic peak extraction is performed based on the difference data to obtain peak data;

[0012] Based on the peak data, amplitude characteristic analysis is performed to obtain the noise dynamic threshold;

[0013] Pulse detection is performed based on the difference data and the noise dynamic threshold to obtain the action position information of the action potential in the high-throughput neural signal.

[0014] In some embodiments, the differential processing of the high-throughput neural signal to obtain difference data includes the following steps:

[0015] The high-throughput neural signal is segmented using a sliding window to obtain window data;

[0016] Subtract the window data corresponding to the first sliding window from the window data corresponding to the second sliding window to obtain the difference data, wherein the first sliding window and the second sliding window are adjacent sliding windows, and the first sliding window precedes the second sliding window in time.

[0017] In some embodiments, the step of dynamically extracting peak values ​​based on the difference data to obtain peak data includes the following steps:

[0018] The difference data is averaged to obtain the average difference data.

[0019] Based on the first adjustable coefficient, noise compensation is performed on the average difference data to obtain the negative peak baseline;

[0020] The difference data is compared with the negative peak baseline to obtain a first comparison result;

[0021] When the first comparison result is that the negative absolute value of the difference data is less than the negative peak baseline, the negative absolute value of the difference data is determined as the peak data.

[0022] In some embodiments, the step of performing amplitude characteristic analysis based on the peak data to obtain the noise dynamic threshold includes the following steps:

[0023] The peak data is averaged to obtain the noise baseline.

[0024] The noise baseline is adjusted based on the second adjustable coefficient to obtain the dynamic noise threshold.

[0025] In some embodiments, the step of performing impulse detection based on the difference data and the noise dynamic threshold to obtain the action potential's action location information in the high-throughput neural signal includes the following steps:

[0026] The difference data is compared with the noise dynamic threshold to obtain a second comparison result;

[0027] When the second comparison result is that the difference data is less than the noise dynamic threshold, the second sliding window corresponding to the difference data is determined as the action potential's action location information in the high-throughput neural signal.

[0028] In some embodiments, the step of performing event-driven adaptive variable-rate compression on the high-throughput neural signal based on the action location information to obtain compressed data includes the following steps:

[0029] Based on the action position information, determine whether the window data contains the action potential, and obtain the determination result;

[0030] If the determination result indicates that the action potential exists, then the window data is compressed with fidelity according to the first preset sampling rate to obtain the first compressed signal;

[0031] When the determination result is that the action potential does not exist, the window data is downsampled and compressed according to the second preset sampling rate to obtain a second compressed signal, wherein the second preset sampling rate is less than the first preset sampling rate;

[0032] Compressed data is obtained based on the first compression signal and the second compression signal.

[0033] To achieve the above objectives, another aspect of this application proposes an adaptive compression system based on differential filtering of neural signal action potential detection, the system comprising:

[0034] A neural signal sampling module is used to acquire high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, and the potential signal includes action potential and local field potential;

[0035] An action potential difference detection module is used to perform difference calculation and detection on the high-throughput neural signal based on the high-frequency narrow pulse abrupt change characteristics, so as to obtain the action position information of the action potential in the high-throughput neural signal;

[0036] A variable rate compression module is used to perform event-driven adaptive variable rate compression on the high-throughput neural signal based on the action position information to obtain compressed data.

[0037] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0038] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0039] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0040] The embodiments of this application include at least the following beneficial effects: This application provides an adaptive compression method and related equipment for action potential detection of neural signals based on difference filtering. This scheme acquires high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, including action potentials and local field potentials. Based on the high-frequency narrow pulse mutation characteristics, difference calculations are performed on the high-throughput neural signals to detect the action potentials, obtaining the action position information within the high-throughput neural signals. By proposing difference filtering based on the characteristics of action potentials and local field potentials to replace high-throughput filtering, accurate detection of action potentials can be achieved. Simultaneously, by performing event-driven adaptive variable-rate compression on the high-throughput neural signals based on the action position information, compressed data is obtained. Dynamic variable-rate compression based on the characteristics of the two signals can be performed, greatly reducing resource consumption and enabling hardware implementation of parallel compression of thousands of channels of neural signals. Attached Figure Description

[0041] Figure 1 This is a flowchart of the adaptive compression method for action potential detection of neural signals based on difference filtering, provided in an embodiment of this application.

[0042] Figure 2 This is a schematic diagram of the waveform of the original neural signal provided in the embodiments of this application;

[0043] Figure 3 This is a flowchart of the difference calculation detection provided in the embodiments of this application;

[0044] Figure 4 This is a flowchart of an adaptive compression method for action potential detection of neural signals based on difference filtering, provided in another embodiment of this application;

[0045] Figure 5 This is a schematic diagram of the structure of the adaptive compression system for action potential detection of neural signals based on difference filtering, provided in an embodiment of this application.

[0046] Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of systems and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0049] Before providing a detailed description of the embodiments of this application, some related technologies involved in the embodiments of this application will be described first.

[0050] Among related technologies, implantable brain-computer interfaces (BCIs) directly implant microelectrode arrays into brain tissue, enabling the recording of electrical signals from the cerebral cortex at the neural scale. This is currently the most direct and reliable method for acquiring neural data with high temporal resolution and high spatial precision. This type of technology plays a crucial role in revealing the dynamic changes in neural activity and the mechanisms of information transmission, providing a solid foundation for cognitive neuroscience research and the treatment of neurological diseases.

[0051] In recent years, with the continuous advancement of microelectronics technology, biocompatible materials and packaging technology, implantable brain-computer interface systems have shown two significant development trends: one is the continuous increase in the number of channels, from the initial few channels to the current ability to acquire synchronous neural signals from hundreds or even thousands of channels.

[0052] Secondly, there is a growing demand for system miniaturization and full implantability. Research shows that neural activity recordings during free-movement states provide more ecologically valid brain data compared to restrained experimental conditions. This trend requires brain-computer interface systems to possess wireless communication capabilities, low-power computing, and the ability to operate stably for extended periods.

[0053] Building on this foundation, research on brain-computer interfaces is gradually expanding from primates to marine organisms. On the one hand, marine animals such as dolphins, whales, and octopuses have highly developed nervous systems, and their complex social behaviors, spatial navigation, and acoustic communication mechanisms are of great significance for understanding the adaptive evolution of the brain. By collecting neural signals from these marine organisms through brain-computer interfaces, researchers can reveal their unique perception and cognitive patterns more deeply.

[0054] On the other hand, brain-computer interface research in the marine environment also has unique value in engineering and application. For example, if the neural activity of marine mammals in their natural habitat can be recorded and analyzed in real time, it will help to establish a more natural behavioral experimental paradigm and promote information interaction between humans and marine animals.

[0055] Furthermore, in marine exploration applications, if brain-computer interfaces can be used to enhance the behavioral control or information perception capabilities of marine organisms, it will provide new ideas for deep-sea exploration and underwater communication.

[0056] However, while increasing the number of channels helps to obtain more comprehensive brain region information, it also leads to an exponential increase in the amount of data generated per second by the system, thus placing higher demands on data storage, real-time transmission, and back-end processing capabilities.

[0057] In summary, with the development of high-throughput neural signal acquisition and wireless communication technologies, brain-computer interfaces are increasingly promising in marine biology research and practical applications. However, they also face key challenges such as adaptability to complex environments, low-power signal processing, and efficient data compression.

[0058] In view of this, this application provides an adaptive compression method and related equipment for action potential detection of neural signals based on difference filtering. This method acquires high-throughput neural signals, including at least one potential signal, which includes action potentials and local field potentials. Based on the high-frequency narrow pulse mutation characteristics, difference calculations are performed on the high-throughput neural signals to detect the action potentials and obtain their position information within the signals. By proposing difference filtering based on the characteristics of action potentials and local field potentials to replace high-throughput filtering, accurate detection of action potentials can be achieved. Simultaneously, by performing event-driven adaptive variable-rate compression on the high-throughput neural signals based on the action position information, compressed data is obtained. Dynamic variable-rate compression based on the characteristics of both signals significantly reduces resource consumption and enables hardware implementation of parallel compression of thousands of neural signals, facilitating long-term monitoring of the free behavior of marine organisms.

[0059] The adaptive compression method based on differential filtering of neural signal action potential detection provided in this application relates to the field of biosignal detection technology. This adaptive compression method can be applied to terminals, servers, or software running on either a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the adaptive compression method based on differential filtering of neural signal action potential detection, but is not limited to the above forms.

[0060] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0061] Figure 1 This is an optional flowchart of the adaptive compression method for action potential detection of neural signals based on difference filtering provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S103.

[0062] Step S101: Acquire high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, and the potential signal includes action potential and local field potential.

[0063] Step S102: Based on the high-frequency narrow pulse mutation characteristics, perform difference calculation to detect the high-throughput neural signal and obtain the action position information of the action potential in the high-throughput neural signal.

[0064] Step S103: Perform event-driven adaptive variable rate compression on the high-throughput neural signal based on the action position information to obtain compressed data.

[0065] In this embodiment, the high-throughput neural signals acquired by the implanted brain-computer interface system mainly consist of action potentials (APs) and local field potentials (LFPs), such as... Figure 2 As shown, the action potentials exhibit extremely narrow pulses with high frequency components, and different neurons have different firing intervals. The local field potentials are low-frequency fluctuating signals, and the fluctuations of local field potentials collected from test points that are close to each other show similarities and significant correlations.

[0066] Specifically, in the natural state, the neural activity of marine organisms changes dramatically with their behavior (such as swimming, hunting, and communicating). By implanting a brain-computer interface system into the cerebral cortex of marine organisms, high-throughput neural signals containing a mixture of action potentials and local field potentials can be collected.

[0067] It should be noted that in order to ensure the integrity of the action potential during neural signal sampling, a high sampling rate needs to be set. Typically, to acquire the AP signal, each electrode needs to be sampled at a Nyquist rate of 30 kHz with at least 10-bit resolution.

[0068] When the number of integrated electrodes exceeds 1000, the system will generate data up to 300 Mbps. This leads to oversampling during the action potential interval, resulting in a large amount of redundant data. However, the communication bandwidth of wireless systems is limited. To resolve the contradiction between data volume and communication bandwidth, preprocessing and compressing redundant data has become an essential problem that must be solved for the development of brain-computer interfaces in the field of marine biological monitoring.

[0069] After acquiring high-throughput neural signals from thousands of channels, unlike schemes that rely on high-pass filters to extract action potentials, this embodiment proposes difference filtering based on the distinguishing characteristics between action potentials and local field potentials to replace digital filters such as high-pass and low-pass filters for separating high-throughput neural signals.

[0070] Specifically, current neural data compression schemes require a high resource ratio for hardware implementation using digital filters. When the number of channels reaches thousands, hardware implementation becomes difficult, especially in low-power implantable systems.

[0071] To improve system energy efficiency and reduce hardware implementation complexity, this embodiment, while sacrificing some compression ratio, utilizes the high-frequency narrow pulse characteristics of short-duration amplitude changes before and after the action potential, and the low-frequency fluctuation characteristics of local field potential, to effectively suppress background signal interference through a narrow window differential method. This can be simplified to detecting the action potential through differential calculation.

[0072] Specifically, the difference operation detection measures the dynamic change of the potential signal by calculating the difference between high-throughput neural signals over a certain time interval within a local time. Then, using the overall fluctuation level of the high-throughput neural signal, a threshold for evaluating amplitude abrupt changes is dynamically calculated.

[0073] If the change in the difference response exceeds the threshold, it indicates that there may be a rapidly rising or falling pulse, which is consistent with the high-frequency narrow pulse characteristics of short amplitude abrupt changes before and after the action potential. The action potential's position information in the original signal can be obtained using the above comparison method. This action position information is used to identify the time of occurrence of the action potential in the time series data of high-throughput neural signals.

[0074] like Figure 3 As shown, the difference operation detection includes processes such as difference calculation, peak value calculation, noise baseline calculation, and threshold detection. The action potential localization can be completed using only simple operations such as addition, subtraction, shifting, Boolean AND comparison. It can not only achieve accurate detection of action potential, but also avoid the frequent use of resource-intensive multiplication and accumulation units in the detection stage. This significantly reduces the hardware overhead of the signal preprocessing stage, provides optimized data input for subsequent efficient compression, and greatly reduces resource consumption.

[0075] It should be noted that during the difference calculation detection process, the difference calculation and other steps for each channel are independent, and the thresholds calculated for each channel are different.

[0076] In the compression stage, an event-driven adaptive sampling strategy is adopted based on the presence and distribution characteristics of action potentials. When an action potential is detected, a high sampling rate is maintained to preserve the neural spike waveform completely, providing high-fidelity data for neuron classification and pattern recognition. When no action potential is found in the detection window, the sampling rate is reduced by a certain compression ratio to remove redundant information and achieve high compression, thereby realizing adaptive variable rate compression.

[0077] For example, during the compression phase, if the marine organism's neural activity is active, the complete neural signals are recorded with high fidelity to ensure the accuracy of action potential detection and reconstruction. Conversely, when the marine organism is at rest, efficient compression significantly reduces the data volume, allowing more effective information to be recorded within a limited bandwidth. This method achieves a low-power, high-efficiency hardware compression scheme for high-throughput neural signals while maintaining the accuracy of action potential detection and the integrity of neural signal characteristics. It enables the parallel compression of thousands of neural signals, facilitating long-term monitoring of the free behavior of marine organisms.

[0078] In some embodiments, step S102 may include, but is not limited to, steps S201 to S204.

[0079] Step S201: Perform differential processing on the high-throughput neural signal to obtain differential data.

[0080] Step S202: Dynamic peak extraction is performed based on the difference data to obtain peak data.

[0081] Step S203: Analyze the amplitude characteristics based on the peak data to obtain the noise dynamic threshold.

[0082] Step S204: Based on the difference data and noise dynamic threshold, pulse detection is performed to obtain the action position information of the action potential in the high-throughput neural signal.

[0083] In this embodiment, differential processing, when processing high-throughput neural signals, addresses the transient characteristics of action potentials, which undergo drastic voltage changes within a very short time. The difference amplifies these transient features. Conversely, local field potentials in high-throughput neural signals exhibit a more gradual change, resulting in similar voltage values ​​at different sampling points. The difference after subtraction is small, approaching zero, and can be effectively suppressed. By leveraging the order-of-magnitude difference in the rates of change of these two potential signals in the time domain, differential processing of the original high-throughput neural signal yields differential data that is easily identifiable and distinguishable between action potentials and local field potentials.

[0084] After obtaining the difference data, peak data with significant amplitude that may be caused by action potentials are initially screened out from the difference data.

[0085] Specifically, by performing statistical analysis on the difference data, a peak baseline can be obtained to describe the overall amplitude level of a set of difference data. This peak baseline can further distinguish between peaks caused by action potentials and peaks caused by local field potential fluctuations. By filtering the difference data whose amplitude abrupt changes do not meet the characteristics of high-frequency narrow pulse abrupt changes of action potentials through the peak baseline, only the difference data exceeding the peak baseline is taken as peak data, so as to reduce the amount of data in subsequent processing.

[0086] It should be noted that the overall amplitude level of the difference data will fluctuate due to the influence of different detection channels and biological states. If a fixed peak baseline is used, when the signal weakens as a whole, for example due to the decrease in channel sensitivity, the fixed standard may not be able to detect the real action potential. Therefore, when determining the peak baseline, the median or mean is dynamically calculated based on the real-time acquired difference data to adaptively determine the peak baseline.

[0087] Furthermore, since some peak data also contain background noise, it is necessary to perform amplitude characteristic analysis on the identified peak data, quantify the noise intensity using the peak data, and determine a noise dynamic threshold that distinguishes noise from action potential.

[0088] Finally, each difference data obtained is compared with the noise dynamic threshold of the corresponding channel for impulse detection. By traversing the entire data stream, each difference data is compared. If the amplitude of the difference data exceeds the noise dynamic threshold, it can be considered that there is an action potential in the original high-throughput neural signal. Thus, the action position information of the action potential in the high-throughput neural signal can be determined according to the time corresponding to the difference data.

[0089] In some embodiments, step S201 may include, but is not limited to, steps S301 to S302.

[0090] Step S301: Use a sliding window to segment the high-throughput neural signals to obtain window data.

[0091] Step S302: Subtract the window data corresponding to the first sliding window from the window data corresponding to the second sliding window to obtain the difference data. The first sliding window and the second sliding window are adjacent sliding windows, and the first sliding window precedes the second sliding window in time.

[0092] In this embodiment, a sliding window is used to segment high-throughput neural signals. By dividing continuous high-throughput neural signals into a series of continuous short time intervals in time, window data is obtained.

[0093] Specifically, high-throughput neural signals can have data volumes as high as 300 Mbps. Directly storing or processing these signals would require excessive computational resources, making it impractical. Therefore, a fixed-length window is used to perform sliding segmentation on the high-throughput neural signals. By processing the data in segments, the instantaneous computational load is significantly reduced.

[0094] For example, such as Figure 3As shown, the data is segmented within the window with a step size of 15 sampling points, dividing the high-throughput neural signals into several groups of window data. For example, the high-throughput neural signals from sampling point 1 to sampling point 15 are one group of window data, and the high-throughput neural signals from sampling point 16 to sampling point 30 are another group of window data.

[0095] Next, based on the temporal relationship between the sliding windows, two adjacent sliding windows are used as action potential detection windows for difference calculation, wherein the first sliding window precedes the second sliding window in time. The data from the first and second sliding windows are aligned, and the corresponding data from the first and second sliding windows are subtracted to obtain the difference data.

[0096] For example, refer to Figure 3 Taking windows Time1, Time2 and Time3 as examples, window Time1 includes sampling points 1 to 15, window Time2 includes sampling points 16 to 30, and window Time3 includes sampling points 31 to 45.

[0097] Window Time1 and window Time2 are adjacent time windows. Window Time1 precedes window Time2 in time. Therefore, window Time1 is the first sliding window and window Time2 is the second sliding window.

[0098] The first difference data is obtained by subtracting the data of sampling point 1 from the data of sampling point 16, the second difference data is obtained by subtracting the data of sampling point 2 from the data of sampling point 17, and so on, to obtain a set of 15 difference data from the first 30 sampling points.

[0099] Window Time2 and window Time3 are adjacent time windows. Window Time2 precedes window Time3 in time. Therefore, window Time2 is the first sliding window and window Time3 is the second sliding window.

[0100] Similarly, subtracting the data from the data at sampling point 16 from the data at sampling point 31 yields the first difference data, subtracting the data at sampling point 17 from the data at sampling point 32 yields the second difference data, and so on, to obtain 15 difference data from sampling points 16 to 45.

[0101] In some embodiments, step S202 may include, but is not limited to, steps S401 to S404.

[0102] Step S401: Perform difference averaging on the difference data to obtain the difference average data.

[0103] Step S402: Based on the first adjustable coefficient, noise compensation is performed on the difference average data to obtain the negative peak baseline.

[0104] Step S403: Compare the difference data with the negative peak baseline to obtain the first comparison result.

[0105] Step S404: When the first comparison result is that the negative absolute value of the difference data is less than the negative peak baseline, the negative absolute value of the difference data is determined as the peak data.

[0106] In this embodiment, the difference data is averaged to obtain a statistical estimate of the overall amplitude level of the difference data within the current action potential detection window, providing a basis for setting a reasonable screening baseline.

[0107] Specifically, in order to eliminate the difference in direction between positive and negative differences, this embodiment first takes the absolute value of the difference data to unify the direction of amplitude change, and then uses the sum of absolute values ​​to calculate the average difference data.

[0108] Next, as Figure 3 As shown in the peak calculation, the peak baseline is calculated, the negative of the average difference data is taken and multiplied by the first adjustable coefficient for noise compensation, and the negative peak baseline is obtained.

[0109] Specifically, the first adjustable coefficient is used to compensate for channel noise. During the same neural signal acquisition process, the noise levels of different channels (i.e., different microelectrodes) are not significantly different; therefore, the same first adjustable coefficient can be used to compensate for all channels. The difference average data is multiplied by the first adjustable coefficient corresponding to each channel to compensate for noise in the difference average data.

[0110] Understandably, for neural signal acquisition in different scenarios, since the noise levels vary greatly in different scenarios, the first adjustable coefficient of all channels can be adjusted according to the noise level of different scenarios.

[0111] For example, the first adjustable coefficient can be set to 1.6. If the channel noise level is high or the signal quality level is poor, the first adjustable coefficient can be further increased to 1.8, 2.0, etc.

[0112] Furthermore, before comparing the difference data with the negative peak baseline, the absolute value of the difference data is first negative to obtain a negative absolute value, and then compared with the negative peak baseline to obtain the first comparison result.

[0113] If the first comparison result shows that the negative absolute value of the difference data is less than the negative peak baseline, it means that the amplitude change of the difference data exceeds the threshold set by the negative peak baseline, which can be considered to meet the characteristics of pulse mutation. In this case, the negative absolute value of the difference data is determined as the peak data.

[0114] In some embodiments, step S203 may include, but is not limited to, steps S501 to S502.

[0115] Step S501: Perform peak averaging based on the peak data to obtain the noise baseline.

[0116] Step S502: Adjust the noise baseline based on the second adjustable coefficient to obtain the noise dynamic threshold.

[0117] In this embodiment, the peak data determined in step S202 includes not only the actual action potential but also a certain amount of background noise. In order to eliminate the interference of background noise and avoid misjudgment, this embodiment accumulates the peak data and takes the average value to evaluate the noise baseline of the channel in the current window.

[0118] In complex environments, noise distribution may vary, and a fixed threshold cannot achieve optimal detection performance in all scenarios. Therefore, a second adjustable coefficient is used to adjust the noise baseline threshold based on the application scenario. The impact of noise background on neural signals acquired in different scenarios varies, and this adjustable coefficient can be set to adjust the noise baseline threshold.

[0119] For example, in an ideal experimental environment, the background noise level is low, while in a marine environment, environmental interference is significant. Therefore, by setting a higher second adjustable coefficient, such as 3, the detection threshold of action potential can be increased, which greatly ensures that the neural signal detected as action potential is real and effective.

[0120] In some embodiments, step S204 may include, but is not limited to, steps S601 to S602.

[0121] Step S601: Compare the difference data with the noise dynamic threshold to obtain a second comparison result.

[0122] Step S602: When the second comparison result is that the difference data is less than the noise dynamic threshold, the second sliding window corresponding to the difference data is determined as the action potential's action location information in the high-throughput neural signal.

[0123] In this embodiment, the difference data within the action potential detection window is compared with the calculated noise dynamic threshold to determine the action potential. The difference of a real action potential at the time of difference will significantly exceed the noise dynamic threshold. Therefore, by obtaining the second comparison result through comparison calculation, the presence of action potential in the high-throughput neural signal within the action potential detection window can be detected.

[0124] When the second comparison result shows that the difference data is less than the noise dynamic threshold, the second sliding window corresponding to the difference data is determined as the action potential's position information in the high-throughput neural signal. Since each difference data point is obtained by subtracting a corresponding sampling point in the first sliding window from a sampling point within the second sliding window, when a difference data point is determined to be a valid pulse, it signifies a dramatic signal change occurring within the time interval of the second sliding window. The position of this difference data within the second sliding window is the action potential's position information in the high-throughput neural signal.

[0125] In some embodiments, step S103 may include, but is not limited to, steps S701 to S704.

[0126] Step S701: Determine whether there is an action potential in the window data based on the action position information, and obtain the determination result.

[0127] In step S702, if the determination result is that an action potential exists, the window data is compressed with fidelity according to the first preset sampling rate to obtain the first compressed signal.

[0128] Step S703: When the determination result is that there is no action potential, the window data is downsampled and compressed according to the second preset sampling rate to obtain the second compressed signal, wherein the second preset sampling rate is less than the first preset sampling rate.

[0129] Step S704: Obtain compressed data based on the first compression signal and the second compression signal.

[0130] In this embodiment, the sampling strategy is dynamically adjusted based on the action potential detection results. While ensuring action potential detection accuracy and signal reconstruction quality, parallel compression of thousands of neural signals can be achieved. Although some trade-offs are made compared to some high compression ratio algorithms, this embodiment achieves a balance between compression performance and hardware feasibility, meeting the practical application requirements of implantable neural interfaces under low power consumption, limited bandwidth, and high throughput conditions.

[0131] Specifically, the action location information clearly identifies which window data contains valid action potential events. Therefore, by traversing the time index and action location information of the window data, it is determined whether an action potential exists in the window data, and the result is obtained. If the time index of the window data is within the set of action location information, the result is "action potential exists"; otherwise, the result is "action potential does not exist".

[0132] Based on different judgment results, event-driven adaptive variable rate compression dynamically adjusts the sampling strategy according to the presence or absence of action potentials (APs), and adjusts the sampling rate according to the characteristics of the two potential signals.

[0133] Specifically, when an action potential is detected, in order to fully preserve the temporal characteristics of the spike signal and ensure the accuracy of subsequent neuronal classification and behavioral pattern recognition of marine organisms, the window data is compressed with fidelity at a first preset sampling rate to obtain a first compressed signal.

[0134] For example, assuming the sampling rate of the original high-throughput neural signal is 30 kSa / s, and the window data includes data from 15 sampling points, in order to ensure the high fidelity of the action potential, the first preset sampling rate can be set to 30 kSa / s to achieve lossless compression. Its core objective is to preserve as much detail as possible of the original waveform. At this time, the focus of compression is not on obtaining an extremely high compression ratio, but on ensuring high progress of the waveform during reconstruction. Therefore, the 15 data points in the window are directly transmitted to obtain the first compressed signal.

[0135] When no action potential is detected, it means that there are only low-frequency local field potentials (LFP) in the window data. For low-frequency local field potentials (LFP), maintaining the original sampling rate would introduce a lot of redundant data. Therefore, under the condition of satisfying the Nyquist sampling theorem, equal-interval downsampling is performed by reducing the sampling rate by a second preset sampling rate. By retaining one sampling point every multiple sampling points, redundant data can be effectively removed and compression efficiency can be improved.

[0136] For example, when an LFP signal is detected, the sampling rate of the original high-throughput neural signal is reduced by the compression ratio. If the compression ratio is 30, the second preset sampling rate is 1 kSa / s, and if the compression ratio is 10, the second preset sampling rate is 3 kSa / s.

[0137] Finally, the first and second compressed signals are integrated according to the time sequence of the original high-throughput neural signals to obtain compressed data.

[0138] The following is a detailed description and explanation of the solutions in the embodiments of the present invention, using specific application examples:

[0139] like Figure 4As shown, another embodiment of this application proposes an adaptive compression method for action potential detection of high-throughput neural signals based on difference filtering. This method can be implemented in hardware. On the one hand, it abandons digital filters in the action potential detection process and adopts a difference operation detection strategy to achieve rapid location of action potentials without filters. This avoids the resource and power consumption overhead caused by large-scale multiplication and accumulation operations brought about by filtering, thereby significantly reducing hardware complexity. On the other hand, it adopts an event-driven compression strategy: it maintains high-fidelity data compression when action potentials are detected and adopts a high compression ratio mode when no action potentials are detected.

[0140] Specifically, action potential-related feature information is first extracted through difference calculation, including difference features, peak amplitude, noise baseline, and threshold estimation.

[0141] When calculating the difference, a step size of 15 is used. The 16th value is subtracted from the 1st value. This process is repeated to obtain 15 differences between the first 30 points. These 15 values ​​are used as a window of data. The negative of the average absolute value of the window data is taken and denoted as mean_window.

[0142] Next, peak value determination and baseline calculation are performed. The negative absolute value of the original window data is compared with 1.6 * mean_window. If the value is less than 1.6 * mean_window, it is considered a peak value. The peak value is accumulated and stored in a register, and the average value is taken as the baseline value of that channel.

[0143] The threshold is obtained by multiplying the coefficient a=3 (adjustable) by the channel baseline value, and the difference within the window is compared with the threshold. The original signal can obtain the position information of the action potential by detecting the threshold. The difference is marked as Spike, which provides a reliable detection basis for subsequent adaptive variable rate compression.

[0144] After obtaining the position of the Spike in the raw data, the sampling strategy is dynamically adjusted according to the presence of the action potential. When there is an action potential in the detection window, the original sampling rate is maintained to ensure the high fidelity of the signal, thereby completely preserving the spike signal. When there is no action potential in the detection window, equal-interval downsampling is enabled, and the data in the window is compressed and transmitted with a compression ratio of about 15 times, effectively removing redundant data.

[0145] The above methods not only significantly reduce the amount of data while ensuring the accuracy of action potential detection, thus effectively supporting the efficient transmission and storage of high-channel-count neural signals in hardware, but also provide an efficient and feasible solution for hardware implementation in applications such as wireless implantable brain-computer interfaces. This effectively alleviates the contradiction between the massive amount of data brought about by the continuous increase in the number of channels and the limited communication bandwidth and battery capacity, and helps to achieve miniaturization, low-power operation and long-term stable operation of brain-computer interface systems. This allows for the recording of neural activity in the free behavior state of marine organisms, providing more ecologically valid brain data for researchers to establish more natural behavioral experimental paradigms, and laying a technical foundation for the application of brain-computer interfaces in marine biology research and practice.

[0146] It should be noted that the adaptive compression method based on differential filtering neural signal action potential detection provided in this application can be applied not only to signal detection in marine organisms but also to primates, as long as neural signals can be acquired through a brain-computer interface. This application is merely illustrative and does not impose any specific limitations.

[0147] Reference Figure 5 This application also provides an adaptive compression system based on differential filtering of neural signal action potential detection, which can implement the above method. The system includes:

[0148] A neural signal sampling module is used to acquire high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, and the potential signal includes action potential and local field potential.

[0149] The action potential difference detection module is used to perform difference calculations on high-throughput neural signals based on the high-frequency narrow pulse abrupt change characteristics to obtain the action position information of the action potential in the high-throughput neural signal.

[0150] The variable rate compression module is used to perform event-driven adaptive variable rate compression on high-throughput neural signals based on action position information to obtain compressed data.

[0151] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0152] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0153] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0154] Reference Figure 6 , Figure 6 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0155] The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0156] The memory 902 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 using the methods described in the embodiments of this application.

[0157] The input / output interface 903 is used to implement information input and output.

[0158] The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0159] Bus 905 transmits information between various components of the device, such as processor 901, memory 902, input / output interface 903, and communication interface 904.

[0160] The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0161] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0162] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0163] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0164] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0165] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0166] The adaptive compression method and related equipment for action potential detection based on differential filtering of neural signals provided in this application embodiment acquire high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, including action potentials and local field potentials; perform differential calculation on the high-throughput neural signals based on the high-frequency narrow pulse mutation characteristics to obtain the action position information of the action potentials in the high-throughput neural signals; and perform event-driven adaptive variable-rate compression of the high-throughput neural signals according to the action position information to obtain compressed data. This application embodiment proposes differential filtering based on the characteristics of action potentials and local field potentials to replace high-throughput filtering, which can achieve accurate detection of action potentials, perform dynamic variable-rate compression based on the characteristics of the two signals, and greatly reduce resource consumption, thus completing the hardware implementation of parallel compression of thousands of channels of neural signals.

[0167] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0168] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0169] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0170] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0171] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0172] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0173] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0174] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0175] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0176] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0177] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. An adaptive compression method for action potential detection of neural signals based on difference filtering, characterized in that, The method includes the following steps: Acquiring high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, the potential signal including action potential and local field potential; The high-throughput neural signal is detected by difference operation based on the high-frequency narrow pulse mutation characteristics to obtain the action position information of the action potential in the high-throughput neural signal; Based on the action location information, the high-throughput neural signal is subjected to event-driven adaptive variable rate compression to obtain compressed data; The method of performing difference calculations on the high-throughput neural signal based on the high-frequency narrow pulse mutation characteristics to obtain the action potential's action location information within the high-throughput neural signal includes the following steps: The high-throughput neural signal is subjected to differential processing to obtain difference data; Dynamic peak extraction is performed based on the difference data to obtain peak data; Based on the peak data, amplitude characteristic analysis is performed to obtain the noise dynamic threshold; Pulse detection is performed based on the difference data and the noise dynamic threshold to obtain the action position information of the action potential in the high-throughput neural signal; The differential processing of the high-throughput neural signal to obtain difference data includes the following steps: The high-throughput neural signal is segmented using a sliding window to obtain window data; Subtract the window data corresponding to the first sliding window from the window data corresponding to the second sliding window to obtain the difference data, wherein the first sliding window and the second sliding window are adjacent sliding windows, and the first sliding window precedes the second sliding window in time; The step of dynamically extracting peak values ​​based on the difference data to obtain peak data includes the following steps: The difference data is averaged to obtain the average difference data. Based on the first adjustable coefficient, noise compensation is performed on the average difference data to obtain the negative peak baseline; The difference data is compared with the negative peak baseline to obtain a first comparison result; When the first comparison result is that the negative absolute value of the difference data is less than the negative peak baseline, the negative absolute value of the difference data is determined as the peak data.

2. The method according to claim 1, characterized in that, The step of performing amplitude characteristic analysis based on the peak data to obtain the noise dynamic threshold includes the following steps: The peak data is averaged to obtain the noise baseline. The noise baseline is adjusted based on the second adjustable coefficient to obtain the dynamic noise threshold.

3. The method according to claim 1, characterized in that, The step of performing impulse detection based on the difference data and the noise dynamic threshold to obtain the action potential's action location information in the high-throughput neural signal includes the following steps: The difference data is compared with the noise dynamic threshold to obtain a second comparison result; When the second comparison result is that the difference data is less than the noise dynamic threshold, the second sliding window corresponding to the difference data is determined as the action potential's action location information in the high-throughput neural signal.

4. The method according to claim 1, characterized in that, The step of performing event-driven adaptive variable-rate compression on the high-throughput neural signal based on the action location information to obtain compressed data includes the following steps: Based on the action position information, determine whether the window data contains the action potential, and obtain the determination result; If the determination result indicates that the action potential exists, then the window data is compressed with fidelity according to the first preset sampling rate to obtain the first compressed signal; When the determination result is that the action potential does not exist, the window data is downsampled and compressed according to the second preset sampling rate to obtain a second compressed signal, wherein the second preset sampling rate is less than the first preset sampling rate; Compressed data is obtained based on the first compression signal and the second compression signal.

5. An adaptive compression system based on differential filtering of neural signal action potential detection, characterized in that, The system includes: A neural signal sampling module is used to acquire high-throughput neural signals, wherein the high-throughput neural signals include at least one potential signal, and the potential signal includes action potential and local field potential; An action potential difference detection module is used to perform difference calculation and detection on the high-throughput neural signal based on the high-frequency narrow pulse abrupt change characteristics, so as to obtain the action position information of the action potential in the high-throughput neural signal; A variable rate compression module is used to perform event-driven adaptive variable rate compression on the high-throughput neural signal based on the action position information to obtain compressed data. The method of performing difference calculations on the high-throughput neural signal based on the high-frequency narrow pulse mutation characteristics to obtain the action potential's action location information within the high-throughput neural signal includes the following steps: The high-throughput neural signal is subjected to differential processing to obtain difference data; Dynamic peak extraction is performed based on the difference data to obtain peak data; Based on the peak data, amplitude characteristic analysis is performed to obtain the noise dynamic threshold; Pulse detection is performed based on the difference data and the noise dynamic threshold to obtain the action position information of the action potential in the high-throughput neural signal; The differential processing of the high-throughput neural signal to obtain difference data includes the following steps: The high-throughput neural signal is segmented using a sliding window to obtain window data; Subtract the window data corresponding to the first sliding window from the window data corresponding to the second sliding window to obtain the difference data, wherein the first sliding window and the second sliding window are adjacent sliding windows, and the first sliding window precedes the second sliding window in time; The step of dynamically extracting peak values ​​based on the difference data to obtain peak data includes the following steps: The difference data is averaged to obtain the average difference data. Based on the first adjustable coefficient, noise compensation is performed on the average difference data to obtain the negative peak baseline; The difference data is compared with the negative peak baseline to obtain a first comparison result; When the first comparison result is that the negative absolute value of the difference data is less than the negative peak baseline, the negative absolute value of the difference data is determined as the peak data.

6. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 4.

Images