A signal processing method and system combining multi-layer diffusion model with ICA
By combining the multi-layer diffusion model and ICA, the problem of unsatisfactory timing signal separation effect in the prior art is solved, efficient denoising and separation of complex signals is achieved, and purer and more accurate signal components are extracted.
Patent Information
- Application Number
- CN202411882089.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-19
- Publication Date
- 2025-04-04
AI Technical Summary
The prior art uses unsatisfactory signal separation when dealing with complex noise or interference timing signals, and using diffusion models alone may not be able to completely separate all independent components.
Combining the multi-layer diffusion model and ICA, by constructing a multi-layer diffusion model, each layer corresponds to different particle size levels, the original signal is denoised layer by layer, and the denoised signal is separated by ICA, and finally the separated signals of each layer are fused to obtain a pure signal.
It improves the accuracy and robustness of signal separation, and can extract purer and more accurate signal components, which is suitable for denoising and feature extraction in complex signal environments.
Smart Images
Figure CN119760414A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of signal processing, and in particular to a signal processing method and system combining a multi-layer diffusion model with ICA. Background Art
[0002] Time series signal separation is an important issue in the field of signal processing, especially in the analysis of biomedical signal processing (such as EEG signal processing, ECG signal processing, and EMG signal processing). Currently, independent component analysis (ICA) is a commonly used method for time series signal separation. It performs linear transformation on the signal to maximize the independence of each component, thereby achieving preliminary separation of the signal. However, the ICA method often does not achieve ideal separation results when processing time series signals containing complex noise or interference.
[0003] In recent years, deep learning-based diffusion models, such as DDPM (Denoising Diffusion Probabilistic Models), have made significant progress in signal denoising and enhancement. These models achieve data denoising and enhancement by gradually adding noise to the data and then learning the process of recovering the original data from the noise. However, using diffusion models alone for time series signal separation also has certain limitations, such as the inability to completely separate all independent components. Summary of the invention
[0004] The main purpose of the present invention is to solve the problem that the time series signal separation effect in the prior art is not ideal. A signal processing method combining a multi-layer diffusion model and ICA comprises the following steps: Construct a multi-layer diffusion model, where each layer corresponds to a different granularity level; In each layer of the diffusion model, the original signal with added noise is subjected to denoising using the diffusion model to obtain a denoised signal; Applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components; The independent signal components separated from each layer are fused to obtain a pure signal.
[0005] As a preferred technical solution, a multi-layer diffusion model is constructed, wherein each layer corresponds to a different granularity level, including: A multi-layer diffusion model is constructed, and the number of layers is determined to be N=3. The granularity levels corresponding to each layer are coarse-grained, medium-grained, and fine-grained.
[0006] As a preferred technical solution, in each layer of the diffusion model, the original signal with added noise is subjected to denoising using the diffusion model to obtain the denoised signal, including: In the i-th layer (i=1,2,...,N), the original signal is converted into a noise form that is easy to process by gradually adding noise. Let the original signal be x and the signal after adding noise be x 1 (noisy) , then the noise adding process can be expressed as: Among them, σi is the standard deviation of the noise in the i-th layer, and n is the random noise.
[0007] (3) In the i-th layer, the inverse generation process of the diffusion model is used to gradually remove the noise and restore the denoised signal x i (denoised) The reverse generation process can be expressed as: Among them, fi is the denoising function of the i-th layer.
[0008] As a preferred technical solution, the applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components includes: The noise-reduced signal x i (denoised) As input to ICA, construct the mixed signal matrix Ai , and solve the inverse transformation matrix Wi , so that the separated signal matrix S = Wi · Ai The signals in are independent of each other.
[0009] As a preferred technical solution, the method for fusing the independent signal components separated from each layer is weighted averaging or maximum a posteriori probability estimation.
[0010] The second aspect of the present invention provides a time series signal separation iterative optimization system combining a back-diffusion model and ICA, comprising: Building a unit,constructing a multi-layer diffusion model, where each layer corresponds to a different granularity level; A noise reduction processing unit, in each layer of the diffusion model, performs noise reduction processing on the original signal with added noise using the diffusion model to obtain a noise-reduced signal; A signal separation unit, applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components; The fusion unit fuses the independent signal components separated from each layer to obtain a pure signal.
[0011] A third aspect of the present invention provides an electronic device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the electronic device executes the signal processing method combining the multi-layer diffusion model and ICA as described above.
[0012] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the signal processing method combining the multi-layer diffusion model with ICA as described above.
[0013] The present invention has the following beneficial effects: (1) By combining the multi-layer diffusion model with ICA, we can fully utilize the denoising capability of the diffusion model at different granularity levels and the advantages of ICA in signal separation to achieve effective processing and analysis of complex signals.
[0014] (2) The accuracy and robustness of signal separation are improved, and purer and more accurate signal components can be extracted.
[0015] (3) It has broad application prospects and can be applied to biomedical signal processing, communication signal processing, audio signal processing and other fields. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Picture 1 A flow chart of a signal processing method combining a multi-layer diffusion model and ICA provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0017] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0018] For ease of understanding, the specific process of the embodiment of the present invention is described below. Picture 1 The first embodiment of the signal processing method combining the multi-layer diffusion model and ICA in the embodiment of the present invention includes: Construct a multi-layer diffusion model, where each layer corresponds to a different granularity level; In each layer of the diffusion model, the original signal with added noise is subjected to denoising using the diffusion model to obtain a denoised signal; Applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components; The independent signal components separated from each layer are fused to obtain a pure signal.
[0019] This method constructs a multi-layer diffusion model to process the original signal layer by layer. Each layer corresponds to a different granularity level, and the signal is gradually denoised from coarse to fine, and independent component analysis (ICA) is combined to achieve signal separation. Finally, the multi-layer results are fused to obtain a pure target signal. This method is particularly suitable for denoising and feature extraction in complex signal environments.
[0020] Step 1: Build a multi-layer diffusion model The entire model is divided into multiple layers, each corresponding to a different granularity level. The granularity levels progress from coarse to fine.
[0021] Each layer of the model defines the granularity by controlling the size of the noise: Coarse-grained: Adds larger noise, suitable for capturing the global characteristics of the signal.
[0022] Medium granularity: Adds moderate noise to extract the main structure in the signal.
[0023] Fine-grained: Add less noise to preserve the fine information of the signal.
[0024] Step 2: Signal noise reduction layer by layer Add noise: In each layer, noise of varying strengths is gradually added to the input signal to bring it into a more processable form.
[0025] Noise reduction processing: The diffusion model is used to denoise the noisy signal and gradually restore the original main components of the signal. The denoising process extracts the corresponding information based on the characteristics of the signal at different granularity levels.
[0026] Step 3: Signal Separation The denoised signals of each layer are input into the independent component analysis (ICA) module.
[0027] ICA analyzes the denoised signal and separates its independent components: Each independent component represents a source or specific pattern of the signal.
[0028] The separated independent components provide basic data for subsequent signal fusion.
[0029] Step 4: Fusion of separation results Multi-layer signal fusion: The individual components separated in all layers are combined to generate the final clean signal.
[0030] Fusion method: The weighted average method can be used to fuse the signal components of different layers after weighting them according to their importance.
[0031] Alternatively, the optimal combination can be selected through statistical methods (such as maximum a posteriori probability estimation) to ensure signal integrity and quality.
[0032] The details are as follows: Construct a multi-layer diffusion model, in which each layer corresponds to a different granularity level; construct a multi-layer diffusion model, determine the number of layers N=3, and the granularity levels corresponding to each layer are coarse-grained, medium-grained and fine-grained respectively.
[0033] In each layer of the diffusion model, the original signal with added noise is subjected to denoising using the diffusion model to obtain a denoised signal; In the i-th layer (i=1,2,...,N), the original signal is converted into a noise form that is easy to process by gradually adding noise. Let the original signal be x and the signal after adding noise be x 1 (noisy) , then the noise adding process can be expressed as: Among them, σi is the standard deviation of the noise in the i-th layer, and n is the random noise.
[0034] (3) In the i-th layer, the inverse generation process of the diffusion model is used to gradually remove the noise and restore the denoised signal x i (denoised) The reverse generation process can be expressed as: Among them, fi is the denoising function of the i-th layer.
[0035] Apply ICA to the de-noised signal to separate the signal and obtain the separated independent signal components; x i (denoised) As input to ICA, construct the mixed signal matrix Ai , and solve the inverse transformation matrix Wi , so that the separated signal matrix S = Wi · Ai The signals in are independent of each other.
[0036] The independent signal components separated from each layer are fused to obtain a pure signal. The method for fusing the independent signal components separated from each layer is weighted average or maximum a posteriori probability estimation.
[0037] The more specific implementation methods of the present invention are as follows: (1) Construct a multi-layer diffusion model and determine the number of layers N = 3. The granularity levels corresponding to each layer are coarse-grained, medium-grained, and fine-grained.
[0038] (2) In the first layer (coarse-grained layer), the original signal is converted into a noise form that is easy to process by gradually adding noise. Suppose the original signal is the EEG signal, and the signal after adding noise is x 1 (noisy) .
[0039] (3) In the first layer, the reverse generation process of the diffusion model is used to gradually remove the noise and restore the denoised signal. x 1 (denoised) .
[0040] (4) The noise-reduced signal x 1 (denoised) As input to ICA, construct the mixed signal matrix A 1, and solve the inverse transformation matrix W 1. Get the separated signal matrix S 1.
[0041] (5) Repeat steps (2)-(4) to perform noise reduction and signal separation in the second layer (medium-grained layer) and the third layer (fine-grained layer), respectively.
[0042] (6) The signals separated from each layer are fused to obtain the final pure EEG signal. The fusion strategy can adopt the weighted average method.
[0043] (7) The signal processing method of the present invention is applied to other types of biomedical signals (such as ECG, EMG, etc.) for verification, and the performance difference is compared with the prior art.
[0044] Taking EEG (electroencephalogram), ECG (electrocardiogram) and EMG (electromyogram) signals as examples, the processing of these signals in the multi-layer diffusion model can be explained as follows: First layer (coarse-grained layer): In this layer, the model may treat the EEG, ECG, and EMG signals as a whole and try to remove the common noise among them. This can be achieved by adding noise in the forward process of the diffusion model and removing it in the backward process. Since this layer is coarse-grained, it may not be able to completely separate the independent components in the EEG, ECG, and EMG signals, but it can remove some obvious noise.
[0045] Second layer (medium granularity layer): In this layer, the model may begin to try to separate out some of the individual components in the EEG, ECG, and EMG signals. This can be done by applying signal separation techniques such as ICA. Since this layer is more fine-grained than the first layer, it is able to capture some of the more subtle features in the signal and try to separate those out.
[0046] The third layer (fine-grained layer): In this layer, the model may further refine the processing of EEG, ECG, and EMG signals. It may try to separate more specific features in the signal, such as different frequency bands of activity in EEG signals, different heartbeat cycles in ECG signals, etc. This layer may require the use of more complex models and techniques to capture and separate these more subtle signal components.
[0047] In each layer, the diffusion model plays a role in noise reduction, while ICA is used for signal separation. As the number of layers increases, the model processes the signal more and more finely, and can capture more details and features. Finally, the signals separated by each layer are fused to obtain the final pure signal.
[0048] It should be noted that the above is just a simplified example. The actual multi-layer diffusion model may contain more layers and more complex processing. In addition, the specific number of layers and the processing in each layer may depend on factors such as the type and complexity of the signal and the processing target.
[0049] In general, the multi-layer diffusion model can fully utilize the noise reduction ability of the diffusion model at different granularity levels and the advantages of signal separation techniques such as ICA by representing and processing signals at different granularity levels, thereby achieving effective processing and analysis of complex signals.
[0050] The above describes the signal processing method combining the multi-layer diffusion model and ICA in the embodiment of the present invention. The following describes a time series signal separation iterative optimization system combining the diffusion model and ICA in the embodiment of the present invention: Building a unit,constructing a multi-layer diffusion model, where each layer corresponds to a different granularity level; A noise reduction processing unit, in each layer of the diffusion model, performs noise reduction processing on the original signal with added noise using the diffusion model to obtain a noise-reduced signal; A signal separation unit, applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components; The fusion unit fuses the independent signal components separated from each layer to obtain a pure signal.
[0051] An embodiment of the present invention also provides an electronic device, which may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) (for example, one or more processors) and memories, and one or more storage media for storing applications or data (for example, one or more mass storage devices). Among them, the memory and the storage medium may be short-term storage or permanent storage. The program stored in the storage medium may include one or more modules, each module may include a series of instruction operations in the electronic device. Furthermore, the processor may be configured to communicate with the storage medium and execute a series of instruction operations in the storage medium on the electronic device.
[0052] The electronic device may also include one or more power supplies, one or more wired or wireless network interfaces, one or more input and output interfaces, and / or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that the electronic device structure in this embodiment does not constitute a limitation on the electronic device, and may include more or fewer components, or combine certain components, or arrange components differently.
[0053] An embodiment of the present invention provides a structure of an electronic device, which may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) (for example, one or more processors) and memories, and one or more storage media for storing applications or data (for example, one or more mass storage devices). Among them, the memory and the storage medium can be short-term storage or permanent storage. The program stored in the storage medium may include one or more modules, each module may include a series of instruction operations in the electronic device. Furthermore, the processor can be configured to communicate with the storage medium and execute a series of instruction operations in the storage medium on the electronic device.
[0054] The electronic device may also include one or more power supplies, one or more wired or wireless network interfaces, one or more input and output interfaces, and / or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that the electronic device structure does not constitute a limitation on the electronic device, and may include more or less components than the aforementioned, or combine certain components, or arrange the components differently.
[0055] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of a signal processing method combining a multilayer diffusion model with ICA.
[0056] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.
[0057] If the integrated unit is implemented in the form of 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 the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.
[0058] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A signal processing method combining a multi-layer diffusion model and ICA, characterized in that: The following steps are involved: Construct a multi-layer diffusion model, where each layer corresponds to a different granularity level; In each layer of the diffusion model, the original signal with added noise is subjected to denoising using the diffusion model to obtain a denoised signal; Applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components; The independent signal components separated from each layer are fused to obtain a pure signal.
2. The signal processing method combining a multi-layer diffusion model and ICA according to claim 1, characterized in that: The multi-layer diffusion model is constructed, wherein each layer corresponds to a different granularity level, including: A multi-layer diffusion model is constructed, and the number of layers is determined to be N=3. The granularity levels corresponding to each layer are coarse-grained, medium-grained, and fine-grained.
3. The signal processing method combining a multi-layer diffusion model and ICA according to claim 1, characterized in that: In each layer of the diffusion model, the original signal with added noise is subjected to denoising using the diffusion model to obtain a denoised signal, including: In the i-th layer (i=1,2,...,N), the original signal is converted into a noise form that is easy to process by gradually adding noise. Let the original signal be x and the signal after adding noise be x 1 (noisy) , then the noise adding process can be expressed as: Among them, σi is the standard deviation of the noise in the i-th layer, and n is the random noise. (3) In the i-th layer, the inverse generation process of the diffusion model is used to gradually remove the noise and restore the denoised signal x i (denoised) The reverse generation process can be expressed as: Among them, fi is the denoising function of the i-th layer.
4. The signal processing method combining a multi-layer diffusion model and ICA according to claim 3 is characterized in that: The applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components includes: The noise-reduced signal x i (denoised) As input to ICA, construct the mixed signal matrix Ai , and solve the inverse transformation matrix Wi , so that the separated signal matrix Si = Wi · Ai The signals in are independent of each other.
5. The signal processing method combining a multi-layer diffusion model and ICA according to claim 1, characterized in that: The method for fusing the independent signal components separated from each layer is weighted averaging or maximum a posteriori probability estimation.
6. A time series signal separation iterative optimization system combining back diffusion model and ICA, characterized in that: The system comprises: Building a unit,constructing a multi-layer diffusion model, where each layer corresponds to a different granularity level; A noise reduction processing unit, in each layer of the diffusion model, performs noise reduction processing on the original signal with added noise using the diffusion model to obtain a noise-reduced signal; A signal separation unit, applying ICA to perform signal separation on the noise-reduced signal to obtain separated independent signal components; The fusion unit fuses the independent signal components separated from each layer to obtain a pure signal.
7. An electronic device, comprising a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute each step of the signal processing method combining a multi-layer diffusion model with ICA as described in any one of claims 1 to 5.
8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the signal processing method combining the multi-layer diffusion model with ICA as claimed in any one of claims 1 to 5 are implemented.