A reading method, system and computer medium of a multi-channel neural signal acquisition system

Abstract

The application provides a reading method and system of a multi-channel neural signal acquisition system and a computer medium, multiple neural signal acquisition channels are divided into channel groups and multiple reading layers are formed. The reading layer adopts a nested structure, so that the jump between the selected channels can cross the channel group and / or the lower reading layer, thereby reducing the jump cost. In any reading period, only the selected channels are continuously read, and the unselected channels are detected by an independent detection module. According to the reading condition and the effectiveness, the matching degree is evaluated, and the selected channels or the reading layer division in the next period are dynamically adjusted. Through limited channel selection, the high signal value channel is continuously read, and the resource utilization efficiency is improved. The algorithm power consumption is reduced without digitizing the unselected channel. The scheme enhances the adaptive ability of the system to signal changes, and can adapt to the neural signal fluctuation without manual intervention, thereby improving the flexibility and robustness of the system.

Description

Technical Field

[0001] This invention relates to the field of neural signal acquisition, and in particular to a reading method, system, and computer medium for a multi-channel neural signal acquisition system. Background Technology

[0002] High-throughput neural signal acquisition systems play a central role in the field of neural interfaces, primarily capturing weak bioelectrical signals through electrode arrays to achieve continuous monitoring and data readout of neuronal activity. As systems evolve towards large-scale integration and low power consumption, the control method of the readout circuitry becomes a key factor determining the system's resource utilization efficiency.

[0003] However, existing readout circuit configurations generally lack flexibility, making it difficult to optimize resource allocation in real time based on dynamic signal changes. Traditional fixed correspondences or preset polling mechanisms cannot effectively distinguish the actual value of electrode signals, leading to redundant occupation of system resources on ineffective channels and reducing overall utilization efficiency. In scenarios with multiple concurrent signals, fixed readout structures can easily result in limited time resolution or loss of critical information. Furthermore, when faced with non-deterministic physical connection structures, static mapping strategies cannot perceive fluctuations in the actual connection state, making it difficult to guarantee readout stability and reliability when connection success rates vary.

[0004] In summary, there is currently no method that can efficiently and accurately read multi-channel neural signal acquisition systems. Summary of the Invention

[0005] In order to overcome the above-mentioned technical defects, the purpose of this invention is to provide a reading method, system and computer medium for a multi-channel neural signal acquisition system.

[0006] This invention discloses a method for reading signals from a multi-channel neural signal acquisition system, comprising the following steps:

[0007] Multiple neural signal acquisition channels are divided into several channel groups, and each channel group selects only a portion of the channels within the group as the objects of continuous reading; the channel groups are configured to: balance the reading load among the channel groups; the channel groups form multiple reading layers; the multiple reading layers are in a nested structure; so that when jumping from one selected channel to the next selected channel, the higher-level reading layer is used, bypassing the intermediate channel groups without signals and / or the lower-level reading layers, thereby reducing the jumping cost;

[0008] Within any given reading cycle, only the selected channel is continuously read; the reading methods include: sequentially scanning and reading the selected channel according to a pre-defined priority configuration; and / or, when the data is transferred from a lower-level reading layer to a higher-level reading layer, simultaneously sampling the selected channel in the lower-level reading layer, and then sequentially scanning and reading it according to the selection status and priority of the higher-level reading layer.

[0009] The unselected channels are tested using an independent detection module, and the validity of their signals is determined.

[0010] Based on the signal readings of the selected channels and the signal validity of the unselected channels, assess the degree of matching between the reading strategy and the signal distribution in the current reading cycle;

[0011] Based on the degree of matching, maintain or adjust the selected channels and / or adjust the division of channel groups and reading layers in the next time period to update the reading strategy.

[0012] This can be understood as dividing multiple neural signal acquisition channels into selected and unselected channels. By selecting a limited number of channels, the system can ensure continuous reading of the selected channels with high signal value, significantly improving the overall utilization efficiency of hardware resources. Simultaneously, through an independently configured detection module, real-time monitoring of all channels is achieved with extremely low computing power without digitizing the unselected channels. This not only ensures high reading efficiency but also greatly reduces computing overhead, providing core support for the miniaturization of acquisition devices. Furthermore, this solution significantly enhances the system's adaptability to dynamic changes in electrode signals, automatically adjusting the reading strategy based on monitoring results. It adapts to fluctuations in the spatial distribution and temporal characteristics of neural signals without static configuration or manual intervention, thus allowing high reading efficiency for multiple neural signal acquisition channels with low computing power while greatly improving the system's flexibility and robustness.

[0013] Furthermore, through a nested read layer architecture, the system transforms linear channel searching into hierarchical addressing logic. This design allows higher-level read layers to directly perceive the activity status of each channel group, enabling it to quickly skip numerous intermediate channel groups or lower levels that lack valid signals when the control signal searches for the target channel. This path optimization mechanism significantly reduces the time consumed by the read control signal in transmitting and hopping between channels, effectively reducing hopping overhead and ensuring the efficient and stable operation of the read control logic in high-throughput scenarios. It also guarantees the timeliness of the read operation.

[0014] This solution also provides flexible readout timing strategies. The sequential scan mode can flexibly schedule resources according to preset priorities; while the simultaneous sampling mode solves the phase deviation problem caused by time differences in polling readout by freezing the electrical signals of the selected channels during hierarchical transmission. This mechanism ensures high coherence of signals from different channels in the time dimension, enabling the system to maintain high readout efficiency while meeting the requirements of neuroscience research for high temporal resolution and high-quality synchronization of multi-channel signals.

[0015] Preferably, the detection module judges the validity of the signal based on feature values, which include at least one of the following: signal amplitude, energy, spectral characteristics, projection domain characteristics, signal-to-noise ratio, neural event occurrence rate, temporal correlation, spatial transformation characteristics, and statistical characteristics processed by dimensionality reduction algorithm; when the feature value is greater than the threshold of the detection module, the signal of the unselected channel is judged to be valid, and the unselected channel is marked as a flag.

[0016] By using various characteristic values ​​of the signal, such as amplitude, energy, spectrum, projection domain, and spatial transformation, as validity criteria, the detection module can quickly identify potential signal sources in a coarse-grained manner and mark the status of unselected channels. This approach avoids costly digital processing of invalid background noise, enabling the system to pre-select valuable candidate channels from complex electrical signals using only simple feature recognition logic. This not only saves significant computational resources but also provides a reliable triggering basis for the precise scheduling of subsequent reading modules.

[0017] Preferably, adjusting the selected channel in the next time period includes:

[0018] One-to-one replacement; replaces the selected channels with no valid signal and / or weak valid signal with the unselected channels marked as marked;

[0019] Global replacement; all neural signal acquisition channels are read, and new channels are selected by evaluating the signal-to-noise ratio and / or electrical signal amplitude of the channels.

[0020] Similarly, this application also provides two-dimensional dynamic update mechanisms, significantly enhancing the system's adaptability to dynamic changes in neural signals. The one-to-one replacement strategy allows the system to perform micro-tuning on individual channels where signals disappear or weaken without interrupting overall acquisition; while the global replacement strategy ensures that the readout window dynamically moves along with spatial changes in the neural firing activity area through periodic re-evaluation. This mechanism adapts to changes in the spatial distribution and temporal characteristics of neural signals without manual intervention, maintaining the system's long-term flexibility.

[0021] Preferably, the threshold of the detection module is configured in at least one of the following ways:

[0022] Use globally unified threshold parameters;

[0023] Adaptive thresholds are used in each channel;

[0024] When the multi-channel neural signal acquisition system is powered on for the first time, the noise level of each channel is sampled, and the threshold of each channel is set based on the noise level.

[0025] Diverse threshold configuration options further enhance detection accuracy and robustness. Fixed thresholds allow users to balance the number of readable targets and accuracy by adjusting the threshold. Adaptive thresholds, for example, dynamically estimate the noise level of each channel through peak extraction circuitry, addressing the issue of large differences in background interference between channels and effectively preventing false triggering or missed detection of weak signals caused by environmental noise. Alternatively, each channel can be sampled directly upon initial power-on, achieving accurate inspection of each channel. This approach minimizes the risk of false detection while maintaining detection sensitivity, achieving a controllable trade-off between performance and power consumption.

[0026] Preferably, the reading of the detection module is configured as follows:

[0027] The multi-channel neural signal acquisition system periodically reads out the marked state of the unselected channels globally or in sections. When at least one unselected channel is detected to be in a marked state, the matching degree is deemed unqualified.

[0028] Alternatively, the unselected channels can be logically combined. When the multi-channel neural signal acquisition system detects that at least one unselected channel is in a marked state, the matching degree is deemed unqualified. The marked states of all unselected channels are then read out in one go to determine the specific channel location.

[0029] Alternatively, when an unselected channel is marked, the corresponding feature parameters are generated, and the feature parameters of all unselected channels are numerically accumulated or physically superimposed; when the total value of the generated feature parameters exceeds the preset threshold, the matching degree is determined to be unqualified, and the marked status of all unselected channels is read out in one go.

[0030] This can be understood as follows: This solution provides three methods for reading the measurement module. The first utilizes existing readout timing for periodic polling, which has the simplest hardware structure and can reliably statistically analyze the global channel status without additional alarm logic circuitry. The second employs an event-driven mechanism, triggering an interrupt to notify the system only when a valid signal appears in an unselected channel, offering extremely high real-time performance and avoiding the overhead of invalid scanning in the absence of a signal. The final method involves numerical accumulation or physical superposition of physical feature parameters and setting thresholds. The system can detect the marker signal of an unselected channel in the analog domain without activating the digital scanning bus, achieving silent monitoring and precise wake-up. Triggering digital positioning only after a valid event is confirmed significantly reduces the instantaneous bandwidth requirements and standby power consumption of the data link, making it more suitable for large-scale neural signal acquisition.

[0031] Preferably, the reading method further includes:

[0032] When the multi-channel neural signal acquisition system is powered on for the first time, all neural signal acquisition channels are read. By evaluating the signal-to-noise ratio and / or electrical signal amplitude of the channels, new channels are selected.

[0033] This can be understood as follows: the full scan mechanism upon initial power-on establishes a global performance benchmark for the system. By scanning and reading all channels during the initialization phase and evaluating their signal-to-noise ratio or electrical signal thresholds, the system can accurately identify the validity of channels. The key advantage of this mechanism is its ability to automatically identify and eliminate bad sectors with no signal caused by poor electrode connections, physical damage, or impedance anomalies, ensuring that subsequent read resources are allocated only to truly valid channels. This initialization scheme makes readout control no longer dependent on fixed physical channel mappings, giving the system strong versatility and enabling it to adapt to various non-deterministic connection relationships or different topologies of electrode arrays.

[0034] Preferably, adjusting the division of the read layer includes at least one of the following methods:

[0035] The number of read layers is fixed, while the channels and channel groups of each read layer are dynamically adjusted.

[0036] Dynamically select the number of reading layers to ensure a balanced number of channels in each reading layer;

[0037] Adjustments are made based on the spatial location of the selected channel, the distribution of historical events, or statistical characteristics.

[0038] The hierarchical dynamic adjustment strategy endows the system with the ability to regulate extreme signal distributions. By dynamically balancing the number of channels within each level, the system can prevent read path congestion caused by excessively dense local signals. The reconstruction logic, which combines spatial location and historical characteristics, allows the hierarchical structure to evolve with actual discharge patterns, thereby further minimizing hopping overhead and demonstrating the solution's excellent scalability and operational efficiency when dealing with ultra-large-scale electrode arrays.

[0039] A second aspect of this application provides a multi-channel neural signal acquisition system for implementing the reading method of the multi-channel neural signal acquisition system as described in any of the foregoing claims, comprising:

[0040] Multiple neural signal acquisition channels;

[0041] The reading module is used to read signals from the selected channel;

[0042] The detection module is used to perform detection steps to detect the validity of the signal when the neural signal acquisition channel reads the signal;

[0043] The control module is used to initiate the reading process of the detection module and dynamically update the reading strategy of the reading module based on the reading results of the detection module and the reading results of the reading module.

[0044] This system provides a concrete hardware platform for the aforementioned reading method. By heterogeneously integrating a high-performance reading circuit with a low-power detection module, a complete neural signal acquisition platform with self-sensing capabilities is constructed. This makes it possible to achieve high-order-of-magnitude real-time monitoring with an extremely small power budget.

[0045] Preferably, the reading module, detection module, and control module are all implemented in analog domain, digital domain, or analog / digital hybrid domain.

[0046] By explicitly supporting multiple implementations in analog, digital, or mixed domains, this technical solution demonstrates exceptional versatility. This means the readout control method is not dependent on specific hardware processes; it can be integrated into the front-end sensor chip or flexibly deployed through the back-end signal processing unit. This flexibility enables its widespread applicability to neural signal acquisition devices of varying sizes, power budgets, and application scenarios.

[0047] A third aspect of this application provides a computer-readable storage medium including instructions that, when executed by a computer, cause the computer to perform a reading method according to a multichannel neural signal acquisition system as described in any of the preceding claims.

[0048] The computer-readable storage medium provided in this application ensures that the aforementioned complex reading method can be stably stored in the form of an algorithm and quickly migrated between different computing platforms. This gives the technical solution excellent software-defined capabilities, allowing for flexible deployment and updates of the acquisition logic according to different research needs or hardware architectures. Attached Figure Description

[0049] Figure 1 A flowchart illustrating the reading method of the multi-channel neural signal acquisition system provided in this application;

[0050] Figure 2 A schematic diagram of the continuous signal reading method in the reading method of the multi-channel neural signal acquisition system provided in this application;

[0051] Figures 3-5 A schematic diagram of the grouping method in the reading method of the multi-channel neural signal acquisition system provided in this application;

[0052] Figures 6-7 A comparison diagram of non-layered and layered reading methods in the reading method of the multi-channel neural signal acquisition system provided in this application;

[0053] Figure 8 A schematic diagram of the architecture for reading feature parameters in the reading method of the multi-channel neural signal acquisition system provided in this application. Detailed Implementation

[0054] The advantages of the present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments.

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0056] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0057] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0058] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0059] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0060] In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the convenience of the description of the invention and have no specific meaning in themselves. Therefore, "module" and "part" can be used interchangeably.

[0061] Please see Figure 1 , Figures 3-5 , Figure 1 A flowchart illustrating the reading method of the multi-channel neural signal acquisition system provided in this application; Figures 3-5 A schematic diagram of the grouping method in the reading method of the multi-channel neural signal acquisition system provided in this application.

[0062] like Figure 1 As shown, this invention discloses a reading method for a multi-channel neural signal acquisition system, comprising the following steps:

[0063] S100. Divide multiple neural signal acquisition channels into several channel groups. Each channel group selects only a portion of the channels within the group as the objects of continuous reading. The channel groups are configured to: balance the reading load among the channel groups; form multiple reading layers from the channel groups; and have a nested structure among the multiple reading layers. This allows for a jump from one selected channel to the next selected channel, bypassing intermediate channel groups without signals and / or lower-level reading layers through higher-level reading layers, thereby reducing jump overhead.

[0064] S200: During any reading cycle, only the selected channel is continuously read;

[0065] S300: The unselected channel is detected by an independent detection module, and the validity of its signal is determined.

[0066] S400. Based on the signal reading status of the selected channel and the signal validity of the unselected channels, evaluate the degree of matching between the reading strategy and the signal distribution in the current reading cycle.

[0067] S500. Based on the matching degree, maintain or adjust the selected channel and / or adjust the division of channel group and reading layer in the next time period to update the reading strategy.

[0068] The underlying principle needs to be explained here: As neural interface systems develop towards higher channel counts and miniaturization, the contradiction between the explosive growth in the number of electrodes and the limited readout circuit resources (such as chip area, power consumption, and bandwidth) is becoming increasingly prominent. Traditional readout methods often employ either a "one-to-one" independent channel or a "many-to-one" fixed polling. However, in the one-to-one mode, since not all electrodes can capture effective discharges at all times, it leads to a serious waste of readout resources; while in the fixed polling mode, when multiple valid signals occur simultaneously, the rigid readout order causes time multiplexing conflicts, resulting in the loss of critical information.

[0069] Therefore, the reading method of the multi-channel neural signal acquisition system provided in this application first divides multiple neural signal acquisition channels into several channel groups, and these channel groups further constitute multiple reading layers with management functions. During actual operation, the system dynamically divides all channels into two categories based on the signal value assessment results: selected channels and unselected channels. Selected channels refer to those currently identified as having high-value neural signals (such as high signal-to-noise ratio, distinctive discharge events). The system allocates valuable reading resources to these channels for continuous and high-precision signal reading. Unselected channels refer to channels where the current signal is inactive or contains only background noise. These channels do not occupy the bandwidth of the main reading link but are silently monitored by an independent detection module.

[0070] For selected channels, the system samples them using a complete readout circuit. This process involves not only sensing the presence or absence of a signal but also accurately reconstructing the signal amplitude, waveform details, and signal-to-noise ratio. This typically requires processing via an analog-to-digital converter (ADC), demanding significant computing power but yielding the most comprehensive information. For unselected channels, the system utilizes a separate detection module for monitoring. Unlike the readout module, the detection module does not transmit continuous analog signals or pass through an ADC; its core can be understood as a lightweight comparator circuit. It only focuses on whether coarse-grained signal characteristics exceed preset thresholds. Because it eliminates the high-precision digitization process (non-ADC architecture), the detection module's hardware complexity and power consumption are significantly lower than the readout module. This allows the system to monitor thousands of unselected channels in real time with almost no increase in power consumption. This not only ensures high readout efficiency but also greatly reduces computing power overhead, providing core support for the miniaturization of acquisition devices.

[0071] Furthermore, the method provided by this solution also evaluates the matching degree between the signal quality of the selected channel and the alarm flag of the unselected channel in real time. This significantly enhances the system's ability to adapt to dynamic changes in electrode signals. It can automatically adjust the reading strategy according to the monitoring results and adapt to the fluctuations in the spatial distribution and temporal characteristics of neural signals without static configuration or manual intervention. Thus, while allowing high reading efficiency for multiple neural signal acquisition channels with low computing power, it greatly improves the system's flexibility and robustness.

[0072] It should be noted that the aforementioned continuous reading method is also not limited.

[0073] Please see Figure 2 , Figure 2 A schematic diagram of the continuous signal reading method in the reading method of the multi-channel neural signal acquisition system provided in this application.

[0074] like Figure 2 As shown, in one possible implementation, continuously reading signals from the selected channel specifically includes:

[0075] The selected channels are scanned and read out sequentially according to the pre-set priority configuration;

[0076] And / or, when passing data from a lower-level read layer to a higher-level read layer, the selected channels in the lower-level read layer are sampled simultaneously, and then scanned and read out sequentially according to the selection status and priority of the higher-level read layer.

[0077] Two flexible readout timing strategies are provided here. The first is to scan and read out all selected channels in the current level one by one according to a pre-set priority configuration (such as signal strength, preset weights of important brain regions, etc.). The advantage of this mode is the flexibility of resource scheduling. The system can dynamically and on demand allocate the limited readout bandwidth to higher priority channels according to the real-time neural activity state, so as to achieve the optimal utilization of hardware resources.

[0078] In high-precision scenarios involving the study of multi-channel neural firing logic (such as synchronous firing or signal propagation delay), the system can employ a "simultaneous sampling, sequential readout" strategy. Specifically, during signal transmission from lower-level to higher-level readout layers, the system first controls all selected channels in the lower level to perform sampling at the same time. Then, using sample-and-hold logic in the circuit, the analog electrical signal at that instant is "frozen." Subsequently, based on the selection state and priority of the higher-level readout layers, the system sequentially transmits and receives these signals, now "frozen" in their respective channels, to the back-end processing unit through the hierarchical architecture. This ensures high temporal coherence of the data acquired from different channels. This allows the system to eliminate phase distortion caused by polling while maintaining high readout efficiency and without increasing the physical link load, thus meeting the stringent requirements of neuroscience research for high temporal resolution and high-quality synchronous acquisition of multi-channel signals.

[0079] The above is a brief description of the concept of the reading method provided in this application. An embodiment will be provided below as an auxiliary illustration to facilitate a further understanding of the grouping and layering strategies by those skilled in the art:

[0080] like Figures 3-5 As shown, multiple neural signal acquisition channels can be grouped according to the principle of load balancing. For example, such as Figure 3 As shown, we set each channel group to select 3 channels for reading (i.e., 1, 2, and 3 in the diagram), resulting in 7 groups, each corresponding to one column of channels. When a column does not have a valid signal, the group is skipped or merged with an adjacent group. At another time, as... Figure 4 As shown, each channel group is also set to read 3 channels. In this case, channel group 1 corresponds to the first column, channel group 2 corresponds to the second and third columns, channel group 3 corresponds to the fourth and fifth columns, and channel group 4 corresponds to the sixth to eighth columns. This can also ensure that each group has 3 selected channels, which satisfies load balancing.

[0081] Of course, the channel group division can also be dynamically adjusted during operation to adapt to changes in signal distribution. For example, such as... Figure 5As shown, each group can also be configured with 4 effective channels, which will not be elaborated further here. Those skilled in the art can design the number of effective channels in each group as needed, and this application does not impose any restrictions here.

[0082] The above explains the grouping approach. In order to achieve fast switching, it is also necessary to design the structural relationship of the reading layer.

[0083] Please see Figures 6-7 , Figures 6-7 A comparison diagram of non-layered and layered reading methods in the reading method of the multi-channel neural signal acquisition system provided in this application.

[0084] like Figures 6-7 As shown, specifically, multiple read layers are nested in a structure such that when switching from one selected channel to the next, the switching overhead is reduced by passing through higher-level read layers, traversing intermediate groups of channels without signals, and / or lower-level read layers.

[0085] This nested design can be understood as a multi-layered structure with stacked layers. Its core purpose is to alter the trajectory of the readout control signal when searching for the target channel. In traditional polling or linear scanning modes, if the system needs to jump from the first channel to the 10,000th channel, the control signal must physically pass through all the invalid channel units in between, resulting in accumulated jump overhead that severely limits the readout rate. However, with the nested readout layer architecture provided in this application, the system transforms linear search into hierarchical addressing logic. When the control signal searches for the next selected channel, the higher-level readout layer can directly perceive the activity status of the lower channel groups. If a channel group does not contain a selected channel, the control logic can directly "jump" over that group at a higher level, thus quickly skipping a large number of intermediate channel areas without valid signals.

[0086] Taking a system with a 100×100 array, i.e., 10,000 channels, as an example, it can be divided into channel groups of 10 channels each, resulting in 1,000 channel groups. These 1,000 channel groups can be further divided into three read layers of 10×10×10. Taking the same jump from the 1st channel to the 10,000th channel as an example, the traditional read method requires 9,999 jumps (the channel control logic is always that the selected channel must meet the following conditions: high priority and all selected channels have been read. After it has finished reading, the completed information is passed to the lower priority channels. For unselected channels, the information about whether there are channels with higher priority that have not been read needs to be passed to the next channel). After the division, it requires 9 jumps within 10 channel groups, 10 jumps to the first level, 10 jumps to the second level, 10 jumps in the third level, 10 jumps back to the second level, and 10 jumps to the first level, plus 10 jumps in the last channel group, which means only 69 jumps are needed.

[0087] This path optimization mechanism significantly reduces the time consumed by the transmission and switching of readout control signals between channels, effectively reducing the clock cycle occupation during the addressing process. In high-throughput scenarios with tens of thousands of channels or more, this solution not only ensures the high efficiency and stability of the readout control logic, but also ensures the real-time and timely performance of neural signal reading by shortening the addressing blind zone time.

[0088] It will be understood by those skilled in the art that the specific implementation of each step is not limited. The possible specific implementations of each step will be described below with reference to the accompanying drawings.

[0089] Firstly, there are no restrictions on the specific reading layer division and adjustment strategies.

[0090] In one possible implementation, adjusting the partitioning of the read layer includes at least one of the following methods:

[0091] By fixing the number of read layers and dynamically adjusting the channels and channel groups within each read layer, the system can adjust the specific channel range or group members managed by each layer based on real-time signal distribution while maintaining a fixed total number of layers. This approach is suitable for scenarios with relatively fixed hardware structures but where active signal areas frequently shift locally.

[0092] The number of read layers is dynamically selected to ensure a balanced number of channels in each layer. This approach can be understood as the system automatically increasing or decreasing the layer depth based on the total number and density of currently selected channels. By dynamically balancing the number of channels within each layer, the retrieval pressure is evenly distributed across each layer, thus preventing read path congestion caused by excessively dense local signals.

[0093] Adjustments are made based on the spatial location of the selected channel, historical event distribution, or statistical characteristics. The hierarchical structure is divided by combining the physical spatial location of the selected channel on the electrode array, historical discharge patterns, or statistical characteristics (such as hotspot distribution of discharge frequency). Specifically, this method is often used in conjunction with the first two methods. Of course, it can also be used alone; this application makes no restrictions on its use.

[0094] The aforementioned hierarchical division and dynamic adjustment strategy endows the system with the ability to adapt to extreme signal distributions. Through real-time optimization of the hierarchical architecture, the system can effectively prevent the generation of readout bottlenecks. This reconstruction logic, which combines spatial location and historical characteristics, allows the hierarchical structure to evolve with actual discharge patterns, thereby further minimizing jump overhead in dynamic environments. This fully demonstrates the excellent scalability and operational efficiency of the proposed solution when facing ultra-large-scale electrode arrays, ensuring the high-performance of the system in large-scale concurrent signal acquisition.

[0095] Secondly, the criteria for determining the detection channel are also unlimited.

[0096] In one possible implementation, the detection module judges the validity of the signal based on feature values, which include at least one of the following: signal amplitude, energy, spectral characteristics, projection domain characteristics, signal-to-noise ratio, neural event occurrence rate, temporal correlation, spatial transformation characteristics, and statistical characteristics processed by dimensionality reduction algorithm. When the feature value is greater than the threshold of the detection module, the signal of the unselected channel is judged to be valid, and the unselected channel is marked as a flag.

[0097] By using multiple feature values ​​such as signal amplitude, energy, projection domain characteristics, and spectrum as validity criteria, the detection module can quickly identify potential signal sources in a coarse-grained manner and mark unselected channels. This approach avoids costly digital processing of invalid background noise, enabling the system to pre-select valuable candidate channels from complex electrical signals using only simple feature recognition logic. This not only saves significant computational resources but also provides a reliable triggering basis for the precise scheduling of subsequent reading modules.

[0098] Furthermore, the threshold setting method for the above feature values ​​is also not limited.

[0099] In one possible implementation, the threshold of the detection module is configured as at least one of the following schemes:

[0100] Use globally unified threshold parameters;

[0101] Adaptive thresholds are used in each channel;

[0102] When the multi-channel neural signal acquisition system is powered on for the first time, the noise level of each channel is sampled, and the threshold of each channel is set based on the noise level.

[0103] Diverse threshold configuration schemes further enhance detection accuracy and robustness. Fixed thresholds allow users to balance the number of readable targets and accuracy by adjusting the threshold, suitable for applications with relatively stable background environments. Adaptive thresholds, for example, dynamically estimate the noise level of each channel through peak extraction circuitry, addressing the issue of large differences in background interference between channels and effectively preventing false triggers or missed detections of weak signals caused by environmental noise. Upon initial power-on, precise inspection of each channel can be achieved by directly sampling the noise level of each channel and setting the threshold. This initial setting based on real-time noise levels allows the threshold to accurately match the physical characteristics of the interface between the specific electrode and neural tissue. Through the flexible application of these schemes (any one or a combination thereof), the system can minimize the risk of false detection while ensuring detection sensitivity, thus achieving a controllable trade-off between performance and power consumption in actual operation.

[0104] In addition, to cope with signal distributions of different densities, this solution provides a flexible selection channel adjustment strategy.

[0105] In one possible implementation, adjusting the selected channel for the next time period includes:

[0106] One-to-one replacement: Channels with no valid signal and / or weak valid signal in the selected channels are replaced with channels marked as "marked" in the unselected channels. This strategy allows the system to perform micro-tuning on individual channels with missing or weakened signals without interrupting overall acquisition. In this way, the system can transfer resources from low-value channels to newly emerging active channels in real time.

[0107] Global replacement: All neural signal acquisition channels are read, and new channels are selected by evaluating the signal-to-noise ratio and / or electrical signal amplitude of each channel. This strategy ensures that the readout "window" can dynamically move to follow the spatial changes of the neural firing activity area through periodic re-evaluation.

[0108] The two-dimensional dynamic update mechanisms provided in this application significantly enhance the system's adaptability to dynamic changes in neural signals. The one-to-one replacement strategy focuses on real-time optimization at the microscopic level, while the global replacement strategy ensures accurate coverage at the macroscopic level. This mechanism can adaptively respond to changes in the spatial distribution and temporal characteristics of neural signals without manual intervention, thus maintaining high reading efficiency for multiple neural signal acquisition channels with relatively low computational power while preserving the system's long-term flexibility, sensitivity, and robustness.

[0109] Understandably, this global replacement mode can also be used for initial power-on to detect channel validity.

[0110] In one possible implementation, the reading method of this application further includes: when the multi-channel neural signal acquisition system is powered on for the first time, reading all neural signal acquisition channels, and selecting the selected channel within the initial time period by evaluating the signal-to-noise ratio and / or electrical signal amplitude of the channel.

[0111] This can be understood as follows: the full scan mechanism upon initial power-on establishes a global performance benchmark for the system. By scanning and reading out all channels during the initialization phase and evaluating their signal-to-noise ratio or electrical signal thresholds, the system can accurately identify the validity of the channels.

[0112] The significance of this mechanism lies in its ability to automatically identify and eliminate bad channels with no signal caused by poor electrode connections, physical damage, or abnormal impedance. In ultra-large-scale electrode arrays, some electrodes may fail to acquire high-quality neural signals due to the fabrication process or implantation environment. Through full-scale detection upon power-up, the system can pre-exclude these invalid channels from the readout list, ensuring that valuable subsequent readout resources are allocated only to real, signal-valuable valid channels, thereby avoiding wasted readout bandwidth.

[0113] This initialization scheme eliminates the reliance on fixed physical channel mapping for readout control, giving the system exceptional versatility. The system can complete the initial resource configuration through a self-evaluation process without prior knowledge of the electrode distribution, enabling it to flexibly adapt to various non-deterministic connection relationships or different topologies of electrode arrays. This significantly improves the deployment efficiency and flexibility of the equipment under different experimental environments and electrode specifications.

[0114] Finally, there are no restrictions on the sensing and reading schemes for the detection module.

[0115] In one possible implementation, the reading of the detection module is configured as follows:

[0116] The first approach is a periodic scanning method: the multi-channel neural signal acquisition system periodically reads out the marked states of unselected channels globally or in sections. When at least one unselected channel is detected as marked, the matching degree is deemed unqualified. This approach utilizes existing readout timing for periodic polling, resulting in the simplest hardware structure and enabling reliable statistics on the global channel states without the need for additional alarm logic circuitry.

[0117] The second approach uses a logical combination triggering method: Unselected channels are logically combined. When the multi-channel neural signal acquisition system detects at least one unselected channel in a marked state, the matching degree is deemed unqualified. The marked states of all unselected channels are then read out collectively to determine the specific channel location. This approach employs an event-driven mechanism, triggering an interrupt to notify the system only when a valid signal is detected in an unselected channel. This provides extremely high real-time performance and effectively avoids the overhead of invalid scanning in the absence of a signal.

[0118] The third approach involves numerically accumulating or physically superimposing the feature parameters: when an unselected channel is marked, the corresponding feature parameters are generated, and the feature parameters of all unselected channels are numerically accumulated or physically superimposed; when the total value of the generated feature parameters exceeds a preset threshold, the matching degree is deemed unqualified, and the marked status of all unselected channels is read out in a centralized manner.

[0119] Please see Figure 8 , Figure 8 A schematic diagram of the architecture for reading feature parameters in the reading method of the multi-channel neural signal acquisition system provided in this application.

[0120] like Figure 8 As shown, exemplarily, this scheme can be as follows: Each neural signal acquisition channel corresponds to a small current source. When the detection module determines that the feature value of an unselected channel exceeds a preset threshold, the channel will lock a marker state (i.e., logic bit 1). This marker signal serves as a control terminal and is directly coupled to the current source. When the marker state is 1, the current source is turned on, generating a small current (i.e., feature parameter) of a preset magnitude; when the marker state is 0, the current source is turned off. All small current source branches of unselected channels are connected to the same analog current convergence bus. The total current on the convergence bus is equal to the sum of the currents generated by all channels in the marker state. The system has an analog comparator at the end of the convergence bus to compare the real-time total current with a preset total current threshold. When the total current is greater than the total current threshold, the matching degree is determined to be unqualified.

[0121] The technological advantage of this design lies in the fact that the system does not need to activate a power-intensive digital scanning bus, nor does it need to digitally poll tens of thousands of unselected channels one by one. It can instantly sense whether one or more channels have generated valid signal events in the analog domain through the superposition effect of physical currents. Furthermore, since it does not involve analog-to-digital conversion (ADC) and large-scale digital logic switching, the system's standby power consumption is compressed to an extremely low level. It can achieve a balance between readout performance and system power consumption while ensuring real-time signal acquisition, making it more suitable for the long-term stable operation of implantable or high-throughput neural interfaces.

[0122] Those skilled in the art will understand that the above embodiments use a preset small current as a characteristic parameter and measure the current to sense the state of the unselected channel. In another possible implementation, other parameters, such as sound waves, can also be used as characteristic parameters. Those skilled in the art can design them as needed, and this application does not impose any limitations.

[0123] A second aspect of this application provides a multi-channel neural signal acquisition system for implementing the reading method of the multi-channel neural signal acquisition system described in any of the preceding claims, comprising:

[0124] Multiple neural signal acquisition channels;

[0125] The reading module is used to read signals from the selected channel;

[0126] The detection module is used to perform detection steps to detect the validity of the signal when the neural signal acquisition channel reads the signal;

[0127] The control module is used to initiate the reading process of the detection module and dynamically update the reading strategy of the reading module based on the reading results of the detection module and the reading results of the reading module.

[0128] This system provides a concrete hardware platform for the aforementioned reading method. By heterogeneously integrating a high-performance reading circuit with a low-power detection module, a complete neural signal acquisition platform with self-sensing capabilities is constructed. This makes it possible to achieve high-order-of-magnitude real-time monitoring with an extremely small power budget.

[0129] It will be understood by those skilled in the art that there are no restrictions on the specific implementation methods of the above modules.

[0130] In one possible implementation, the readout module, detection module, and control module are all implemented in the analog domain, digital domain, or a hybrid analog / digital domain. By explicitly supporting multiple implementation methods in the analog, digital, or hybrid domains, this technical solution demonstrates strong versatility. This means that the readout control method is not dependent on a specific hardware process; it can be integrated into the front-end sensor chip or flexibly deployed through the back-end signal processing unit. This flexibility makes it widely applicable to neural signal acquisition devices of different sizes, with different power budgets, and in various application scenarios.

[0131] A third aspect of this application provides a computer-readable storage medium including instructions that, when executed by a computer, cause the computer to perform a reading method of a multichannel neural signal acquisition system as described above.

[0132] The computer-readable storage medium provided in this application ensures that the aforementioned complex reading method can be stably stored in the form of an algorithm and quickly migrated between different computing platforms. This gives the technical solution excellent software-defined capabilities, allowing for flexible deployment and updates of the acquisition logic according to different research needs or hardware architectures.

[0133] It should be noted that the embodiments of the present invention have better implementability and are not intended to limit the present invention in any way. Any person skilled in the art may use the above-disclosed technical content to change or modify it into equivalent effective embodiments. However, any modifications or equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A method for reading signals from a multi-channel neural signal acquisition system, characterized in that, Includes the following steps: Multiple neural signal acquisition channels are divided into several channel groups, and each channel group selects only a portion of the channels within the group as the objects to be continuously read; the channel groups are configured to balance the reading load among the channel groups. The plurality of channel groups form multiple readout layers; the plurality of readout layers are in a nested structure; so that when jumping from one selected channel to the next selected channel, the jump overhead is reduced by passing through a higher-level readout layer, bypassing intermediate channel groups without signals and / or lower-level readout layers; During any given reading cycle, only the selected channel is continuously read for signals; The reading methods include: sequentially scanning and reading the selected channels according to a pre-set priority configuration; and / or, when the data is transferred from a lower-level reading layer to a higher-level reading layer, simultaneously sampling the selected channels in the lower-level reading layer, and then sequentially scanning and reading them according to the selection status and priority of the higher-level reading layer. The unselected channels are tested using an independent detection module, and the validity of their signals is determined. Based on the signal reading status of the selected channel and the signal validity of the unselected channel, evaluate the degree of matching between the reading strategy and the signal distribution in the current reading cycle; Based on the matching degree, maintain or adjust the selected channel and / or adjust the division of the channel group and the reading layer in the next time period to update the reading strategy.

2. The reading method of the multi-channel neural signal acquisition system as described in claim 1, characterized in that, The detection module determines the validity of the signal based on feature values, which include at least one of the following: signal amplitude, energy, spectral characteristics, projection domain characteristics, signal-to-noise ratio, neural event occurrence rate, temporal correlation, spatial transformation characteristics, and statistical characteristics processed by a dimensionality reduction algorithm. When the feature value is greater than the threshold of the detection module, the unselected channel signal is determined to be valid, and the unselected channel is marked as a flagged state.

3. The reading method of the multi-channel neural signal acquisition system as described in claim 2, characterized in that, The adjustment of the selected channel in the next time period includes: One-to-one replacement; replacing the selected channels with no valid signal and / or weak valid signal with the unselected channels marked as being in a flagged state; Global replacement; all neural signal acquisition channels are read, and new channels are selected by evaluating the signal-to-noise ratio and / or electrical signal amplitude of the channels.

4. The reading method of the multi-channel neural signal acquisition system as described in claim 2, characterized in that, The threshold of the detection module is configured according to at least one of the following schemes: Use globally unified threshold parameters; An adaptive threshold is used in each of the aforementioned channels; When the multi-channel neural signal acquisition system is powered on for the first time, the noise level of each channel is sampled, and a threshold for each channel is set based on the noise level.

5. The reading method of the multi-channel neural signal acquisition system as described in claim 2, characterized in that, The reading of the detection module is configured as follows: The multi-channel neural signal acquisition system periodically reads out the marked state of the unselected channels globally or in sections. When at least one of the unselected channels is detected to be in a marked state, the matching degree is determined to be unqualified. Alternatively, the unselected channels can be logically combined. When the multi-channel neural signal acquisition system detects that at least one of the unselected channels is in a marked state, the matching degree is determined to be unqualified. The marked states of all unselected channels are read out in one go to determine the specific channel position. Alternatively, when the unselected channel is in a marked state, a corresponding feature parameter is generated, and the feature parameters of all the unselected channels are numerically accumulated or physically superimposed; when the total value of the generated feature parameter exceeds a preset threshold, the matching degree is determined to be unqualified, and the marked state of all the unselected channels is read out in a centralized manner.

6. The reading method of the multi-channel neural signal acquisition system as described in claim 1, characterized in that, The reading method also includes: When the multi-channel neural signal acquisition system is powered on for the first time, all neural signal acquisition channels are read. By evaluating the signal-to-noise ratio and / or electrical signal amplitude of the channels, new selected channels are chosen.

7. The reading method of the multi-channel neural signal acquisition system as described in claim 1, characterized in that, Adjusting the division of the read layer includes at least one of the following methods: The number of read layers is fixed, while the channels and channel groups of each read layer are dynamically adjusted. Dynamically select the number of reading layers to ensure a balanced number of channels in each reading layer; Adjustments are made based on the spatial location of the selected channel, the distribution of historical events, or statistical characteristics.

8. A multi-channel neural signal acquisition system, characterized in that, A reading method for implementing the multi-channel neural signal acquisition system as described in any one of claims 1-7 includes: Multiple neural signal acquisition channels; The reading module is used to read signals from the selected channel; The detection module is used to perform detection steps to detect the validity of the signal when the neural signal acquisition channel reads the signal; The control module is used to initiate the reading process of the detection module and dynamically update the reading strategy of the reading module based on the reading results of the detection module and the reading results of the reading module.

9. The multi-channel neural signal acquisition system as described in claim 8, characterized in that, The reading module, the detection module, and the control module are all implemented in analog domain, digital domain, or analog / digital hybrid domain.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes instructions that, when executed by a computer, cause the computer to perform a reading method of the multichannel neural signal acquisition system according to any one of claims 1-7.

Images