Signal processing method and system based on deep neural network

Abstract

The present application relates to the technical field of signal processing, and discloses a signal processing method and system based on a deep neural network. The method comprises: collecting original signal data; performing preliminary processing on the signal data, comprising filtering and denoising; extracting a key feature of the processed signal; and converting the signal into a format suitable for deep learning processing. The present application fuses signal processing technology and a deep learning method, can process complex nonlinear relationships, and is suitable for various signal processing tasks. The performance and accuracy of a model are improved by means of a large amount of data and computing resources. Meanwhile, the burden of manual feature design is relieved, and important information in a signal can be better captured. A deep neural network has strong nonlinear modeling capabilities, can process complex nonlinear relationships, is suitable for various signal processing tasks, can more accurately model and process complex signals compared with a conventional method, has wide applicability, and can play a role in different fields and applications.

Description

A signal processing method and system based on deep neural network

[0001] The present application claims priority to the Chinese patent application No. 202410454369.3, filed on April 16, 2024, and entitled "A signal processing method based on deep neural network", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application relates to the technical field of signal processing, in particular to a signal processing method and system based on deep neural network. BACKGROUND

[0003] Deep neural network is a kind of artificial neural network, which is composed of multiple layers, usually including input layer, hidden layer and output layer, each layer contains multiple neurons, there are connections between neurons of adjacent layers, and each connection has a weight, through training data, adjusting the weight in the network, so that the network can learn the feature representation of the input data, and perform corresponding classification, recognition or prediction tasks.

[0004] Signal processing refers to the process of acquiring, transforming, transmitting, storing and interpreting signals, which can be data from sensors, communication systems, biological bodies and other sources, the goal of signal processing is to extract useful information from these data and analyze and interpret them.

[0005] Deep neural network can be used for feature extraction, classification, regression and other tasks in signal processing, it can use deep learning model to perform speech recognition on speech signal, target detection on image signal, disease diagnosis on biomedical signal, etc., the powerful feature extraction and nonlinear modeling capabilities of deep learning make it perform well in the field of signal processing.

[0006] Generally, because the related signal processing method usually needs to manually design the feature extractor, it depends on the field knowledge and experience, and is not suitable for complex signal scenarios, it may not perform well when dealing with nonlinear and high-dimensional data, and cannot fully exploit the potential information of the data, for some complex signal processing tasks, the related signal processing method may need a lot of manpower and time cost to adjust parameters and optimize algorithms, and the generalization ability may be weak, and the generalization performance may not be good for new signal types or large changes in data sets.

[0007] In summary, a signal processing method based on deep neural network is needed to solve the above problems. SUMMARY

[0008] The present application aims to provide a deep neural network-based signal processing method and system to solve the problems raised in the background.

[0009] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0010] A deep neural network-based signal processing method, which is implemented based on a signal processing system, includes the following steps:

[0011] S1. Collecting raw signal data;

[0012] S2. Preliminary processing of signal data, including filtering and denoising;

[0013] S3. Extracting key features of processed signals;

[0014] S4. Converting signals into a format suitable for deep learning processing;

[0015] S5. Using deep neural networks to classify, identify or predict signals;

[0016] S6. Training deep learning models to optimize their performance;

[0017] S7. Analyzing DSP code to ensure its correctness and efficiency;

[0018] S8. Analyzing power supply noise and designing filters to optimize power quality;

[0019] S9. Monitoring power supply and system temperature to ensure stable power supply and avoid overheating;

[0020] S10. Converting processing results into a suitable output format;

[0021] S11. Controlling the output of results, including display, storage or transmission;

[0022] S12. Providing a user interface to allow parameter configuration and result viewing;

[0023] S13. Monitoring the status of the entire system to ensure its stable and efficient operation.

[0024] A deep neural network-based signal processing system, which includes a data acquisition module, a signal processing module, a machine learning module, a DSP code review module, a power filter optimization module, a power supply and thermal problem detection module, an output processing module, and a system management module;

[0025] The data acquisition module is used for signal acquisition and preprocessing;

[0026] The signal processing module is used for feature extraction and signal conversion;

[0027] The machine learning module is used for deep learning and model training;

[0028] The DSP code review module is used for code inspection and performance evaluation;

[0029] The power filter optimization module is used for power noise analysis and filter design;

[0030] The power supply and heat problem detection module is used for power supply and temperature monitoring;

[0031] The output processing module is used for data format conversion and output control;

[0032] The system management module is used to provide user interface and system monitoring.

[0033] In an exemplary embodiment, the data acquisition module further includes a signal acquisition unit and a preprocessing unit;

[0034] The signal acquisition unit is used to collect raw signals from sensors or other data sources;

[0035] The preprocessing unit is used to perform preliminary processing on the raw signals, including filtering and denoising, to improve data quality.

[0036] In an exemplary embodiment, the signal processing module further includes a feature extraction unit and a signal conversion unit;

[0037] The feature extraction unit is used to extract key features from the preprocessed signals, including frequency and amplitude;

[0038] The signal conversion unit is used to convert the signal into a format suitable for deep learning model processing, including time-frequency domain conversion.

[0039] In an exemplary embodiment, the machine learning module further includes a deep learning unit and a model training unit;

[0040] The deep learning unit is used to perform signal classification, recognition or prediction tasks using deep neural networks;

[0041] The model training unit is used to train and optimize the deep learning model using a training data set.

[0042] In an exemplary embodiment, the DSP code review module further includes a code inspection unit and a performance evaluation unit;

[0043] The code inspection unit is used to perform static analysis on DSP (Digital Signal Processing) code, checking for potential errors and non-standard coding;

[0044] The performance evaluation unit is used to evaluate the execution efficiency and resource consumption of the DSP code to ensure that performance requirements are met.

[0045] In an exemplary embodiment, the power supply filtering optimization module further comprises a power supply noise analysis unit and a filter design unit;

[0046] The power supply noise analysis unit is used to analyze the noise on the power supply line and determine the filtering requirements;

[0047] The filter design unit is used to design appropriate filters to reduce the impact of power supply noise on signal processing.

[0048] In an exemplary embodiment, the power supply and heat problem detection module further comprises a power supply monitoring unit and a temperature monitoring unit;

[0049] The power supply monitoring unit is used to monitor the power supply voltage and current in real time and detect power supply abnormalities;

[0050] The temperature monitoring unit is used to monitor the system temperature to prevent performance degradation or damage caused by overheating.

[0051] In an exemplary embodiment, the output processing module further comprises a data format conversion unit and an output control unit;

[0052] The data format conversion unit is used to convert the processing results into a format suitable for output or display;

[0053] The output control unit is used to control the output mode of the results, including display screen display, storage or transmission.

[0054] In an exemplary embodiment, the system management module further comprises a user interface unit and a system monitoring unit;

[0055] The user interface unit is used to provide a user interaction interface to allow users to configure parameters and view results;

[0056] The system monitoring unit is used to monitor the system state, including hardware state and software running state, to ensure stable operation of the system.

[0057] Compared with the prior art, the application has the advantages that the application combines signal processing technology and deep learning method, fully utilizes the advantages of both, can automatically learn the advanced feature representation of signals, has strong generalization ability, can process complex nonlinear relationships, is suitable for various signal processing tasks, improves the performance and accuracy of the model through a large amount of data and computing resources, reduces the burden of manual feature design, can better capture important information in signals, and the deep neural network has strong nonlinear modeling ability, can process complex nonlinear relationships, is suitable for various signal processing tasks, can more accurately model and process complex signals compared with traditional methods, has wide applicability, and can play a role in different fields and applications.

[0058] The accompanying drawings

[0059] Figure 1 shows the topology of the signal processing system of the present application;

[0060] Figure 2 is a flowchart of a signal processing method based on deep neural networks according to the present application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0062] Embodiment 1

[0063] As shown in Figure 1, a signal processing system is established, which includes a data acquisition module, a signal processing module, a machine learning module, a DSP code review module, a power filter optimization module, a power supply and heat problem detection module, an output processing module, and a system management module.

[0064] Among them, signal acquisition and preprocessing are performed through the data acquisition module; feature extraction and signal conversion are performed through the signal processing module; deep learning and model training are performed through the machine learning module; code inspection and performance evaluation are performed through the DSP code review module; power noise analysis and filter design are performed through the power filter optimization module; power supply and temperature monitoring are performed through the power supply and heat problem detection module; data format conversion and output control are performed through the output processing module; and user interface and system monitoring are provided through the system management module.

[0065] It should be noted that the original signal from the sensor or other data source (such as a heart sound signal sensor or some one-dimensional signal data source, etc.) is collected through the signal acquisition unit; the original signal is preliminarily processed through the preprocessing unit, including filtering and denoising, to improve data quality.

[0066] It should be noted that the key features, including frequency and amplitude, are extracted from the preprocessed signal through the feature extraction unit; the signal is converted to a format suitable for deep learning model processing through the signal conversion unit, including time-frequency domain conversion.

[0067] It should be noted that the signal classification, recognition or prediction tasks are performed through the deep learning unit using deep neural networks; the deep learning model is trained and optimized using the training data set through the model training unit.

[0068] It is worth noting that the static analysis of DSP (Digital Signal Processing) code is performed by the code checking unit to check potential errors and non-standard coding; the performance evaluation unit evaluates the execution efficiency and resource consumption of the DSP code to ensure that the performance requirements are met.

[0069] It is worth noting that the power supply noise analysis unit is used to analyze the noise on the power line to determine the filtering requirements; the filter design unit is used to design appropriate filters to reduce the impact of power supply noise on signal processing.

[0070] It is worth noting that the power supply monitoring unit is used to monitor the power voltage and current in real time to detect power supply abnormalities; the temperature monitoring unit is used to monitor the system temperature to prevent performance degradation or damage caused by overheating.

[0071] It is worth noting that the data format conversion unit is used to convert the processing results into a format suitable for output or display; the output control unit is used to control the output method of the results, including display screen display, storage or transmission.

[0072] It is worth noting that the user interface unit is used to provide a user interaction interface to allow users to configure parameters and view results; the system monitoring unit is used to monitor the system status, including hardware status and software running status, to ensure stable operation of the system. For example, if the original signal obtained is from a heart sound signal sensor, the system monitoring unit monitors the status of the heart sound signal analysis system where the heart sound signal sensor is located to ensure stable operation of the heart sound signal analysis system and improve the accuracy of subsequent heart sound signal analysis results.

[0073] Embodiment 2

[0074] As shown in FIG. 2, in actual application, based on the signal processing method of the above information processing system, specifically, the following steps are included:

[0075] S1. Collecting original signal data:

[0076] S1.1. Configuring sensors or data sources to start data collection tasks;

[0077] S1.2. Using data collection software to establish data collection channels;

[0078] S1.3. Setting data collection parameters, including sampling rate, sampling precision;

[0079] S1.4. Starting the data collection task to start recording the original signal data;

[0080] S2. Preliminary processing of signal data:

[0081] S2.1. Import the raw data into MATLAB;

[0082] S2.2. Perform low-pass filtering to remove high-frequency noise components;

[0083] S2.3. Perform denoising processing on the filtered data, using techniques such as wavelet denoising or moving average;

[0084] S2.4. Ensure that the processed data retains the effective information of the signal while filtering out noise interference;

[0085] S3. Extract key features of the signal:

[0086] S3.1. Use Fast Fourier Transform to extract frequency and amplitude key features from the preprocessed signal;

[0087] S3.2. Determine the parameters for feature extraction, including window length and overlap rate;

[0088] S3.3. Normalize the extracted features to ensure they are within a certain range;

[0089] S3.4. Verify whether the extracted features are representative and perform feature selection or dimensionality reduction processing if necessary;

[0090] S4. Convert the signal into a format suitable for deep learning processing:

[0091] S4.1. According to the input requirements of the deep learning model, arrange the signal data into appropriate data structures including tensors;

[0092] S4.2. Consider the time sequence of signal data and organize it into time series or sequence data;

[0093] S4.3. Standardize or normalize the data to speed up model convergence and improve training effect;

[0094] S4.4. Divide the processed data set into training set, validation set and test set for model training and evaluation;

[0095] S5. Use deep neural networks for signal processing:

[0096] S5.1. Select a recurrent neural network deep learning model architecture;

[0097] S5.2. Configure the network structure of the model, including the number of layers and the number of neurons parameters;

[0098] S5.3. Train the model using the training set and monitor the loss function changes during training;

[0099] S5.4. Evaluate the model performance using the validation set and perform hyperparameter tuning to improve the model accuracy and generalization ability.

[0100] S6. Train the deep learning model to optimize performance:

[0101] S6.1. Divide the dataset and configure the training, validation, and testing proportions;

[0102] S6.2. Adjust the model parameters using the Adam optimizer;

[0103] S6.3. Monitor the training loss and validation loss during the training process to ensure the model performs well on both the training set and validation set;

[0104] S6.4. Select the best combination of model parameters through cross-validation or hyperparameter search techniques;

[0105] S7. Analyze DSP code:

[0106] S7.1. Use Cppcheck to inspect the DSP code and identify potential code defects;

[0107] S7.2. Consult the documentation and specifications of the DSP processor to ensure the code meets hardware requirements and best practices;

[0108] S7.3. Conduct code reviews with team members to identify potential errors and suggest improvements;

[0109] S7.4. Perform code quality assessment, including evaluating code complexity and maintainability metrics;

[0110] S8. Analyze power supply noise and design filters:

[0111] S8.1. Measure the noise level on the power supply line using an oscilloscope or spectrum analyzer instrument;

[0112] S8.2. Analyze the noise spectrum to determine the frequency range that needs to be filtered and the type of filter required;

[0113] S8.3. Design the filter, selecting appropriate filter parameters to meet design requirements;

[0114] S8.4. Verify and optimize the designed filter performance using MATLAB;

[0115] S9. Monitor power supply and system temperature:

[0116] S9.1. Install a voltmeter, ammeter, and connect them to the system power supply lines;

[0117] S9.2. Set monitoring parameters, including alarm thresholds for voltage, current, and temperature;

[0118] S9.3. Monitor power supply voltage, current, and system temperature in real-time, and record historical data;

[0119] S9.4. Trigger alarms and take appropriate emergency measures when abnormal conditions are detected;

[0120] S10. Convert the processing results into suitable output formats:

[0121] S10.1. Use MATLAB to organize the processed results into acceptable output formats;

[0122] S10.2. Convert the results into text, charts, images, or videos according to requirements, for subsequent analysis or presentation;

[0123] S10.3. Ensure the visualization and readability of output data, and perform necessary formatting and decoration;

[0124] S10.4. Verify the completeness and accuracy of output results, ensuring consistency with the original data processing results;

[0125] S11. Control the output of results:

[0126] S11.1. In the output control module, set the output method, including display, storage, or transmission to a remote server;

[0127] S11.2. Configure output parameters including display resolution and storage format according to system requirements and user preferences;

[0128] S11.3. Implement output control to ensure that results are output as expected;

[0129] S11.4. Monitor the output process and handle possible errors or abnormal situations in a timely manner to ensure the stability and reliability of result output;

[0130] S12. Provide user interface:

[0131] S12.1. Develop user interface software to allow users to interact with the system;

[0132] S12.2. Design user interfaces, including graphical interfaces (GUI) or command-line interfaces (CLI), to meet the needs of different users;

[0133] S12.3. Implement parameter configuration functions to allow users to adjust system parameters and algorithm settings;

[0134] S12.4. Provide result viewing function to enable users to conveniently view processing results and analyze them;

[0135] S13. Monitor the status of the entire system:

[0136] S13.1. Configure system monitoring software to monitor the hardware status and software running status of the system in real time;

[0137] S13.2. Set monitoring indicators and thresholds to timely discover system abnormalities or failures;

[0138] S13.3. Implement system status monitoring, regularly check system running logs and alarm information;

[0139] S13.4. When abnormal conditions are found, take necessary measures including restarting the system, repairing software errors to ensure stable and efficient operation of the system.

[0140] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A deep neural network based signal processing method, the method being implemented based on a signal processing system, characterized by, The method comprises the following steps: S1. Collecting raw signal data; S2. Preliminary processing of signal data, including filtering and denoising; S3. Extracting key features of processed signals; S4. Converting signals into a format suitable for deep learning processing; S5. Using deep neural networks to classify, identify or predict signals; S6. Training deep learning models to optimize their performance; S7. Analyzing DSP code to ensure its correctness and efficiency; S8. Analyzing power supply noise and designing filters to optimize power quality; S9. Monitoring power supply and system temperature to ensure stable power supply and avoid overheating; S10. Converting processing results into a suitable output format; S11. Controlling the output of results, including display, storage or transmission; S12. Providing a user interface to allow parameter configuration and result viewing; S13. Monitoring the status of the entire system to ensure its stable and efficient operation.

2. The signal processing method based on deep neural network according to claim 1, characterized in that, The signal processing system comprises a data acquisition module, a signal processing module, a machine learning module, a DSP code review module, a power filter optimization module, a power supply and thermal problem detection module, an output processing module, and a system management module; The data acquisition module is used for signal acquisition and preprocessing; The signal processing module is used for feature extraction and signal conversion; The machine learning module is used for deep learning and model training; The DSP code review module is used for code checking and performance evaluation; The power filter optimization module is used for power supply noise analysis and filter design; The power supply and thermal problem detection module is used for power supply and temperature monitoring; The output processing module is used for data format conversion and output control; The system management module is used to provide a user interface and system monitoring.

3. The signal processing method based on deep neural networks according to claim 2, characterized in that: The data acquisition module further comprises a signal acquisition unit and a preprocessing unit; The signal acquisition unit is used to collect raw signals from sensors; The preprocessing unit is used to preliminarily process the raw signals.

4. The signal processing method based on deep neural networks according to claim 3, characterized in that: The signal processing module further comprises a feature extraction unit and a signal conversion unit; The feature extraction unit is used to extract key features from preprocessed signals; The signal conversion unit is used to convert signals into a format suitable for deep learning model processing.

5. The signal processing method based on deep neural networks according to claim 4, characterized in that: The machine learning module further comprises a deep learning unit and a model training unit; The deep learning unit is used to perform signal classification, identification or prediction tasks using deep neural networks; The model training unit is used to train and optimize deep learning models using training data sets.

6. The signal processing method based on deep neural networks according to claim 5, characterized in that: The DSP code review module further comprises a code checking unit and a performance evaluation unit; The code checking unit is used for static analysis of DSP code; The performance evaluation unit is used to evaluate the execution efficiency and resource consumption of DSP code.

7. The signal processing method based on deep neural network according to claim 6, wherein: the power supply filtering optimization module further comprises a power supply noise analysis unit and a filter design unit; the power supply noise analysis unit is configured to analyze the noise on the power supply line; the filter design unit is configured to design a suitable filter.

8. The signal processing method based on deep neural network according to claim 7, wherein: the power supply and heat problem detection module further comprises a power supply monitoring unit and a temperature monitoring unit; the power supply monitoring unit is configured to monitor the power supply voltage and current in real time; the temperature monitoring unit is configured to monitor the system temperature.

9. The signal processing method based on deep neural network according to claim 8, wherein: the output processing module further comprises a data format conversion unit and an output control unit; the data format conversion unit is configured to convert the processing result into a format suitable for output or display; the output control unit is configured to control the output mode of the result.

10. The signal processing method based on deep neural network according to claim 9, wherein: the system management module further comprises a user interface unit and a system monitoring unit; the user interface unit is configured to provide a user interaction interface; the system monitoring unit is configured to monitor the system status.

11. A deep neural network based signal processing system, characterized by, The signal processing system comprises a data acquisition module, a signal processing module, a machine learning module, a DSP code review module, a power supply filtering optimization module, a power supply and heat problem detection module, an output processing module, and a system management module; the data acquisition module is configured to acquire and preprocess signals; the signal processing module is configured to extract features and convert signals; the machine learning module is configured to perform deep learning and model training; the DSP code review module is configured to check codes and evaluate performance; the power supply filtering optimization module is configured to analyze power supply noise and design filters; the power supply and heat problem detection module is configured to monitor power supply and temperature; the output processing module is configured to convert data formats and control output; the system management module is configured to provide a user interface and system monitoring.

Images