Brain-computer interface data preprocessing method and apparatus
By separating and compressing brain-computer interface signals and utilizing inter-channel correlation and Kalman filtering methods, the problem of low data transmission efficiency in implantable brain-computer interfaces was solved, achieving efficient data compression and reconstruction and improving communication bandwidth.
Patent Information
- Application Number
- CN202310067647.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-01-18
- Publication Date
- 2025-12-30
- Estimated Expiration
- 2043-01-18
AI Technical Summary
Existing implantable brain-computer interface technologies struggle to achieve a high number of channels and high communication efficiency within limited weight and volume, resulting in high data generation rates, heavy transmission pressure, and significant loss of original information due to existing data compression methods.
By separating the action potential signal and the local field potential signal, variable rate compression is performed using the correlation between channels, and data reconstruction is performed by combining the Kalman filtering method, thus achieving efficient compression and reconstruction of multi-channel data.
It effectively reduces data redundancy, improves data transmission efficiency, retains key information, enables the combined reconstruction of high-frequency, low-precision and low-frequency, high-precision data, and enhances communication bandwidth.
Smart Images
Figure CN116522069B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a brain-computer interface data preprocessing method and apparatus. Background Technology
[0002] Currently, the further application of implantable brain-computer interface (BCI) technology is mainly limited by achieving higher energy efficiency and a higher number of channels within a limited weight and volume. A higher number of channels allows for more comprehensive and higher-resolution information acquisition of the target brain region; however, it also brings higher data generation rates and excessively high data transmission pressure. Furthermore, the wireless communication path of implantable devices needs to pass through the skin, which has a significant filtering and attenuation effect on various wireless communication methods, resulting in limited communication bandwidth. Therefore, data preprocessing techniques are needed to identify and compress existing raw data to reduce data redundancy and improve the efficiency of effective data transmission.
[0003] Existing technologies compress local field potentials based on neural signal characteristics. However, in order to acquire action potentials, the local field potentials are oversampled, resulting in redundant local field potential data. At the same time, current technologies mainly use single-channel data compression schemes, which result in high-proportion lossy compression of transmitted data, leading to the loss of a large amount of original information from multiple channels. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a brain-computer interface data preprocessing method and device that can compress data by utilizing the correlation between channels and reconstruct the original data of each channel using the Kalman method.
[0005] To achieve the above objectives, the present invention adopts the following specific technical solution:
[0006] The present invention provides a brain-computer interface data preprocessing method, comprising the following steps:
[0007] S1. Separate the action potential signal and the local field potential signal from the multi-channel raw signals collected by the brain-computer interface system. The action potential signal is transmitted through the action potential channel, and the local field potential signal is transmitted through the local field potential channel.
[0008] S2. Perform variable-rate compression on the local field potential signal in the local field potential channel to compress the local field potential signal while maintaining a high sampling rate of the action potential signal.
[0009] S3. Based on the correlation characteristics between local field potential channels, the local field potential channels are divided into key channels and compressed channels. The local field potential signals of the key channels and compressed channels are compressed to different degrees. Further compression of the local field potential signals is achieved by further downsampling.
[0010] S4. Reconstruct the local field potential signal of the compression channel based on the action potential signal of the action potential channel and the local field potential signal of the key channel to realize the reorganization of the data stream.
[0011] Furthermore, in step S2, the variable-rate compression of the local field potential signal within the local field potential channel includes the following steps:
[0012] S21. The brain-computer interface system maintains a high sampling rate and detects whether action potential signals are present.
[0013] S22. If an action potential signal exists, the complete data is retained without compression. If no action potential signal exists, the local field potential signal is compressed to reduce the sampling rate, while ensuring that the local field potential signal after compression and reduction of the sampling rate still conforms to the Nyquist sampling theorem.
[0014] S23, Insertion rate frame structure flag, used to distinguish signals with different sampling rates.
[0015] Furthermore, in step S3, compressing the local field potential signals of the key channel and the compression channel to different degrees includes the following steps:
[0016] S31. Perform downsampling on the local field potential signal of the key channel to ensure that the actual sampling rate after downsampling meets the Nyquist sampling theorem and the sampling requirements of the local field potential signal.
[0017] S32. Perform a high-ratio downsampling operation on the local field potential signal of the compressed channel so that the actual sampling rate after the high-ratio downsampling no longer satisfies the Nyquist sampling theorem, resulting in undersampling.
[0018] Further, in step S4, reconstructing the local field potential signal of the compression channel based on the action potential signal of the action potential channel and the local field potential signal of the key channel includes the following steps:
[0019] A Kalman filter prediction scheme is adopted, which combines the action potential signal of the action potential channel with high confidence but large data interval and the local field potential signal of the key channel with low confidence but small data interval to supplement the local field potential signal of the compressed channel and realize the reconstruction of the data stream.
[0020] Furthermore, the brain-computer interface system maintains a high sampling rate of 30 kSa / s, and the sampling rate is reduced to 2 kSa / s after local field potential compression.
[0021] Furthermore, the sampling rate after downsampling of the critical channel is 2 kSa / s, and the sampling rate after high-proportion downsampling of the compressed channel is 500 Sa / s.
[0022] The present invention provides a brain-computer interface data preprocessing device, which uses the above-described method for data preprocessing.
[0023] The present invention can achieve the following technical effects:
[0024] This invention utilizes inter-channel correlation to jointly compress multi-channel data, extracting common information and preserving it at a high rate, while combining it with the low-rate original information within each channel to maximize multi-channel data compression. Simultaneously, it employs the Kalman method to combine high-frequency, low-precision data with low-frequency, high-precision data, achieving high-frequency, high-precision reconstruction capabilities. Attached Figure Description
[0025] Figure 1 This is a schematic diagram showing the location of the data preprocessing method provided in the embodiment of the present invention in a brain-computer interface system.
[0026] Figure 2 This is a schematic diagram of the overall process of the data preprocessing method provided in the embodiments of the present invention.
[0027] Figure 3 This is a schematic diagram of the process of variable rate compression within a channel according to an embodiment of the present invention.
[0028] Figure 4 This is a schematic diagram illustrating the technical principle of compressing local field potential based on channel correlation according to an embodiment of the present invention.
[0029] Figure 5 This is a schematic diagram illustrating the process of data reconstruction using a Kalman filter prediction scheme according to an embodiment of the present invention. Detailed Implementation
[0030] In the following description, embodiments of the invention will be described with reference to the accompanying drawings. In the description below, the same modules are denoted by the same reference numerals. Where the same reference numerals are used, their names and functions are also the same. Therefore, their detailed description will not be repeated.
[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.
[0032] Figure 1 The location of the data preprocessing method provided in the embodiments of the present invention in the brain-computer interface system is shown, such as... Figure 1As shown, the high-density electrode array and the conditioning acquisition on-chip system enable real-time acquisition of neural activity and changes in the surrounding electric field environment, followed by analog-to-digital conversion. The data is then aggregated into a single data stream after scanning via an analog-to-digital converter. This data stream contains the results of the aforementioned multi-channel electrode signals scanned in a fixed order. The data stream enters the logic management chip, is buffered and segmented by the data stream management unit, and then preprocessed and compressed by the data preprocessing unit. The compressed data is then reassembled into a unified data stream using a polling method and transmitted via the wireless data transmission unit.
[0033] Figure 2 The overall flow of the data preprocessing method provided in the embodiments of the present invention is shown, such as... Figure 2 As shown, the brain-computer interface data preprocessing method includes the following steps:
[0034] S1. Separate the action potential signal and the local field potential signal from the multi-channel raw signals collected by the brain-computer interface system. The action potential signal is transmitted through the action potential channel, and the local field potential signal is transmitted through the local field potential channel.
[0035] S2. Variable-rate compression is applied to the local field potential signal in the local field potential channel to compress the local field potential signal while maintaining a high sampling rate of the action potential signal.
[0036] Figure 3 The flowchart of intra-channel variable rate compression provided by an embodiment of the present invention is shown, such as... Figure 3 As shown, the system continuously samples at a sampling rate of 30 kSa / s to ensure accurate capture of the position of the action potential signal and high-precision acquisition of its waveform, preserving the integrity of the original action potential signal without compression.
[0037] The non-action potential signal data segment contains local field potential fluctuation information. A continuous sampling rate of 30 kSa / s will generate serious redundant data. In order to eliminate redundant data, the local field potential data segment is compressed by 15 times. After compression, the local field potential sampling rate is 2 kSa / s, which conforms to the Nyquist sampling theorem and ensures that the actual information of the local field potential signal is not lost.
[0038] S3. Based on the correlation characteristics between local field potential channels, the local field potential channels are divided into key channels and compressed channels. The local field potential signals of the key channels and compressed channels are compressed to different degrees. Further compression of the local field potential signals is achieved by further downsampling.
[0039] The correlation between channels is calculated using the Pearson linear correlation coefficient. The specific formula for calculating the correlation coefficient between channels a and b is as follows:
[0040]
[0041] Where n is the number of points in the channel. The channel with the highest correlation coefficient with other channels within the adjacent channel was selected as the key channel.
[0042] Figure 4 This illustrates the technical principle of local field potential compression based on channel correlation provided in an embodiment of the present invention, such as... Figure 4 As shown, the channel is divided into a critical channel and a compressed channel. The critical channel data is downsampled in accordance with the sampling theorem. After downsampling, the actual sampling rate meets the sampling requirements of the local field potential, thus avoiding undersampling. The compressed channel data is downsampled at a high ratio. After downsampling, the actual sampling rate no longer meets the sampling theorem, which means it is undersampling. Therefore, the compressed channel data needs to be reconstructed in conjunction with the critical channel information during the reconstruction process.
[0043] Each channel of the system uses a sampling rate of 30 kSa / s to acquire the original signal to ensure the acquisition of action potentials. After the action potential detection has been achieved, other data can be acquired by downsampling. The original signal is downsampled by 15 times to achieve a sampling rate of 2 kSa / s.
[0044] The compressed channel data is further compressed by 4 times, resulting in an actual sampling rate of 500 Sa / s. This does not meet the requirements of the sampling theorem, but the local field potential has a strong correlation in adjacent channels. The correlation between these channels can be used to reconstruct the undersampled data.
[0045] S4. Reconstruct the local field potential signal of the compression channel based on the action potential signal of the action potential channel and the local field potential signal of the key channel to realize the reorganization of the data stream.
[0046] A Kalman filter prediction scheme is employed, using signals from the action potential channel and the critical channel as input data. The action potential channel data has high confidence but large data intervals, while the critical channel data has low confidence but small data intervals. By combining the two, the data from the compressed channel is supplemented, thereby achieving data reconstruction.
[0047] Figure 5 The process of data reconstruction using a Kalman filter prediction scheme is illustrated, such as... Figure 5 As shown, the key channel data is used as the basis for each channel data, and the original high-precision data of each channel is used as the prediction correction data. Both types of data are simultaneously input into the Kalman filter algorithm. The algorithm outputs the correction parameters for the key channel data. By correcting the original key channel data with these parameters, the final corrected reconstruction results of each channel are obtained.
[0048] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0049] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
[0050] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.
Claims
1. A brain-computer interface data preprocessing method, characterized in that, The method comprises the following steps: S1, separating the action potential signal and the local field potential signal in the multi-channel original signal collected by the brain-computer interface system, the action potential signal being transmitted through an action potential channel, and the local field potential signal being transmitted through a local field potential channel; S2, performing variable-rate compression on the local field potential signal in the local field potential channel, so as to compress the local field potential signal while maintaining the high sampling rate of the action potential signal; S3, based on the correlation characteristics between the local field potential channels, dividing the local field potential channels into key channels and compression channels, and performing compression on the local field potential signals of the key channels and the compression channels to different degrees, and further reducing the sampling rate to further compress the local field potential signals; S4, reconstructing the local field potential signals of the compression channels according to the action potential signals of the action potential channels and the local field potential signals of the key channels, and recombining the data stream; In the step S4, the reconstruction of the local field potential signals of the compression channels according to the action potential signals of the action potential channels and the local field potential signals of the key channels comprises the following steps: A Kalman filter prediction scheme is adopted, the action potential signal of the action potential channel with high confidence but large data interval is combined with the local field potential signal of the key channel with low confidence but small data interval, the local field potential signals of the compression channels are supplemented, and the data stream is recombined.
2. The brain-computer interface data pre-processing method of claim 1, wherein, In the step S2, the variable-rate compression on the local field potential signal in the local field potential channel comprises the following steps: S21, the brain-computer interface system maintains a high sampling rate, and detects whether the action potential signal exists; S22, if the action potential signal exists, the complete data is reserved and no compression operation is performed; if the action potential signal does not exist, the local field potential signal is compressed to reduce the sampling rate, and it is ensured that the local field potential signal after the compression and the reduction of the sampling rate still meets the Nyquist sampling theorem; S23, a rate frame structure mark is inserted, which is used to distinguish signals with different sampling rates.
3. The brain-computer interface data pre-processing method of claim 1, wherein, In the step S3, the different degrees of compression on the local field potential signals of the key channels and the compression channels comprise the following steps: S31, performing a downsampling operation on the local field potential signal of the key channel, so as to ensure that the actual sampling rate after the downsampling meets the Nyquist sampling theorem and the sampling requirement of the local field potential signal; S32, performing a high-proportion downsampling operation on the local field potential signal of the compression channel, so that the actual sampling rate after the high-proportion downsampling no longer meets the Nyquist sampling theorem, and an under-sampling condition is reached.
4. The brain-computer interface data pre-processing method of claim 2, wherein, The sampling rate of the brain-computer interface system maintaining a high sampling rate is 30kSa / s, and the sampling rate of the local field potential after the compression and the reduction of the sampling rate is 2kSa / s.
5. The brain-computer interface data pre-processing method of claim 3, wherein, The sampling rate of the key channel after the downsampling is 2kSa / s, and the sampling rate of the compression channel after the high-proportion downsampling is 500Sa / s.
6. A brain-computer interface data preprocessing apparatus, characterized by comprising: The method is used for data preprocessing. The method is used for data preprocessing.