Intelligent detection system and method for consumer electronic functional devices

Abstract

The present invention relates to the field of electronic product detection technology, and specifically to an intelligent detection system and method for consumer electronic functional devices, including a signal acquisition module, an interference identification module, a signal processing module, a data analysis module, an anomaly detection module, an adaptive learning module, and a central control unit; wherein the signal acquisition module: collects original electrical signals; the interference identification module: identifies and isolates electromagnetic interference and radio frequency interference components in the signal; the signal processing module: extracts and restores the original interfered signal; the data analysis module: determines the functional status of the product; the anomaly detection module: determines whether there is an anomaly in the functional status of the product; and the adaptive learning module: automatically adjusts the signal processing parameters and analysis algorithms. The present invention, through continuous data collection and analysis, not only improves the accuracy of fault diagnosis, but also reduces unexpected downtime through predictive maintenance, providing greater value for consumer electronic manufacturers and users.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic product detection, and in particular to an intelligent detection system and method for consumer electronic functional devices. Background Art

[0002] In the consumer electronics industry, accurate and timely detection of product functional status is crucial. Traditional functional detection methods mainly rely on manual operation or simple automated equipment. These methods are not only inefficient, but also susceptible to environmental interference, such as electromagnetic and radio frequency interference, resulting in inaccurate detection results. In addition, as the functions of consumer electronic products become increasingly complex, traditional detection methods are difficult to cope with diverse detection needs, such as the need to analyze and judge tiny electrical signal changes, capture short-term fault phenomena, etc. These limitations are particularly evident in high-density electronic environments, where mutual interference between devices may mislead test results, leading to incorrect maintenance decisions and unnecessary cost expenditures.

[0003] In order to overcome these challenges, the present invention provides an intelligent detection system and method for consumer electronic functional devices, which aims to improve the accuracy and automation level of detection by integrating advanced signal processing and intelligent algorithms. The prior art lacks solutions that can effectively resist electronic interference, automatically extract key features of device functional status, and perform real-time learning optimization.

[0004] Therefore, developing a system that can automatically adjust detection parameters and update algorithms in real time to cope with changes in complex electronic environments has become a necessary condition for improving the efficiency and reliability of product quality control. Summary of the invention

[0005] Based on the above objectives, the present invention provides a system and method for intelligent detection of consumer electronic functional devices.

[0006] The intelligent detection system of consumer electronic functional devices includes a signal acquisition module, an interference identification module, a signal processing module, a data analysis module, an anomaly detection module, an adaptive learning module and a central control unit; wherein,

[0007] Signal acquisition module: equipped with a high-sensitivity sensor to collect raw electrical signals from the consumer electronic products to be tested;

[0008] Interference identification module: receives the original electrical signal from the signal acquisition module, and uses the signal analysis technology of spectrum analysis to identify and isolate the electromagnetic interference and radio frequency interference components in the signal;

[0009] Signal processing module: receives the output of the interference identification module, uses digital filtering and signal enhancement technology to extract and restore the interfered original signal to a pure functional signal;

[0010] Data analysis module: receives clean signals from the signal processing module and performs advanced data processing, including feature extraction and status recognition, to determine the functional status of the product;

[0011] Anomaly detection module: receives the results of the data analysis module and uses the preset algorithm to determine whether there is an anomaly in the product function status;

[0012] Adaptive learning module: by receiving data from the signal processing module, data analysis module and anomaly detection module, it uses the preset reinforcement learning algorithm to automatically adjust the signal processing parameters and analysis algorithm;

[0013] Central control unit: connects each module through an internal data bus, coordinates data flow and processing tasks between modules, and is configured with a processor and memory. The processor executes the program stored in the memory and manages data flow and algorithm execution.

[0014] Furthermore, the signal acquisition module includes a high-sensitivity sensor unit, a signal adjustment unit and a data conversion unit; wherein,

[0015] High-sensitivity sensor unit: including piezoelectric sensors and photoelectric sensors, used to collect electrical signals generated by tiny physical changes and light changes from the consumer electronic products to be tested with high precision. Piezoelectric sensors can convert mechanical pressure into electrical signals to capture tiny force changes generated during product operation; photoelectric sensors are used to detect and convert light intensity changes into electrical signals to monitor the status of light-sensitive components or display screens of products;

[0016] Signal conditioning unit: Receives the original electrical signal from the high-sensitivity sensor unit and uses a low-noise amplifier and a bandpass filter to perform preliminary signal amplification and noise filtering to optimize the readability and processability of the signal and ensure that only signals within a predetermined frequency range are transmitted to the interference identification module to reduce interference from environmental noise and non-target signals;

[0017] Data conversion unit: includes an analog-to-digital converter, which is used to convert the conditioned analog signal into a digital signal for subsequent digital processing.

[0018] Furthermore, the interference identification module includes a spectrum analysis unit, an interference detection unit and a signal isolation unit; wherein,

[0019] Spectrum analysis unit: Use fast Fourier transform technology to perform spectrum analysis on the received original electrical signal, and convert the time domain electrical signal x(t) into the frequency domain signal X(f). The specific formula is:

[0020] Where x(t) represents the time domain expression of the original electrical signal; X(f) represents the converted frequency domain signal; f represents the frequency in Hertz; e -j2πft It is a complex exponential function, which is used to convert the time domain signal to the frequency domain;

[0021] Interference detection unit: Based on the output of the spectrum analysis unit, a preset interference frequency template is used for comparative analysis. The interference frequency template includes the frequency range of known electromagnetic interference and radio frequency interference. The comparative analysis process formula is:

[0022]

[0023] Where D(f) represents the interference detection result; X(f) is transmitted from the spectrum analysis unit and represents the frequency domain expression of the signal; predefined EMI / RFI ranges refers to the interference frequency range previously defined based on product design and environmental testing;

[0024] Signal isolation unit: receives the interference frequency component identified by the interference detection unit, and uses a digital band-stop filter to remove the interference frequency component from the original signal. Specifically, the band-stop filter is set to suppress the interference frequency range identified in the interference detection unit. The specific operation is expressed as applying the filter function H(f), and the calculation formula is: Y(f)=X(f)·(1-H(f)), where Y(f) represents the filtered signal, that is, the frequency domain signal after removing the interference component; X(f) is the signal transmitted from the interference detection unit; H(f) is the frequency response function of the band-stop filter.

[0025] Furthermore, the signal processing module includes a digital filtering unit and a signal enhancement unit; wherein,

[0026] Digital filter unit: A preset low-pass filter is used to remove high-frequency noise components and retain the low-frequency signal required for product functions. The specific cutoff frequency of the filter is set based on the operating frequency of the product and the predicted interference frequency. The low-pass filter is expressed by the following equation: Y lp (f) = X(f)·H lp (f), where Y lp (f) represents the signal after low-pass filtering, X(f) is the frequency domain signal from the interference identification module, and H lp (f) is the frequency response function of the low-pass filter, which is 1 below the cutoff frequency and rapidly approaches 0 above the cutoff frequency;

[0027] Signal enhancement unit: Use adaptive enhancement algorithm to improve the signal-to-noise ratio and strength of the signal. The enhancement algorithm analyzes the statistical characteristics of the signal, dynamically adjusts the gain coefficient, and optimizes the signal quality. The formula for signal enhancement is: Y enh(t) = g(t)·x filt (t), where Y enh (t) is the enhanced time domain signal, g(t) is the gain coefficient adaptively calculated based on the signal environment, and x filt (t) is the signal converted back to the time domain after being processed by the digital filter unit.

[0028] Furthermore, the data analysis module includes a feature extraction unit and a state recognition unit; wherein,

[0029] Feature extraction unit: used to extract features from pure signals through wavelet transform. Wavelet transform can provide time and frequency information at multiple scales at the same time, and is used for feature extraction of non-stationary signals. Specifically, the signal is decomposed into wavelet coefficients to accurately capture local changes and anomalies in the signal, which will indicate potential failures in product performance indicators.

[0030] State recognition unit: Based on the wavelet transform features provided by the feature extraction unit, the support vector machine (SVM) algorithm is applied for state recognition. The support vector machine will be trained to distinguish different product functional states, including normal operation, failure mode or performance degradation.

[0031] Furthermore, the feature extraction unit includes:

[0032] Signal preprocessing: Before wavelet transform, the clean signal is first preprocessed, including denoising and normalization. Denoising is implemented by an adaptive filter, which automatically adjusts parameters based on the local variance of the signal to minimize the influence of residual noise; normalization ensures that the signal amplitude is within a predetermined range so that the processing result is not affected by the original strength of the signal; the preprocessing formula is:

[0033] Among them, x(t) is the pure input signal; μ x is the mean value of x(t), that is, the average level of the signal during the entire acquisition period; σ x is the standard deviation of x(t); norm (t) is the normalized signal;

[0034] Wavelet transform application: The preprocessed signal x norm (t) will be sent to the wavelet transform processor, and the Daubechies wavelet with good time-frequency localization characteristics is selected to perform multi-scale decomposition on the signal. The decomposed signal is wavelet coefficients, which are used to capture the characteristics of the signal at different scales. The specific formula is expressed as: W x (a,b)=∫x(t)ψ a，b (t)dt, where x(t) is the normalized signal after preprocessing; ψ a，b(t) is the wavelet function adjusted according to the scale parameter a and the translation parameter b; a is the scale parameter; b is the translation parameter; W x (a, b) is the wavelet coefficient at scale a and translation parameter b, providing the frequency and time information of the signal at this scale and position;

[0035] Feature extraction: From the coefficients obtained by wavelet transform, select the coefficients that can represent the signal attributes as features, and apply threshold processing technology to retain only the coefficients exceeding the threshold to eliminate the weak noise that still exists. The specific formula is:

[0036] F={W x (a i , b i )|E(W x (a i , b i ))>θ}, where W x (a i , b i ) is the coefficient obtained by wavelet transform, representing the signal at scale a i and pan b i Characteristics under E(W x (a i , b i )) is the energy of the wavelet coefficient; θ is the energy threshold; F is the final selected feature set.

[0037] Furthermore, the state recognition unit includes:

[0038] Feature preprocessing: The wavelet transform features received from the feature extraction unit are first normalized to ensure that the feature data has a uniform scale when performing SVM classification;

[0039] SVM model training: Use historical data sets to train SVM classifiers. The features of classification labels in the historical data sets include normal operation, failure mode, and performance degradation. Classifier training includes selecting appropriate kernel functions, including linear kernels, polynomial kernels, or radial basis function kernels, and adjusting the parameters of the kernel function, including the penalty parameter C and the kernel parameter p. The training process formula is expressed as:

[0040]

[0041] Among them, w is the normal vector of the hyperplane, which determines the direction of the classification hyperplane; k is the bias term, which together with the normal vector determines the position of the hyperplane; ξ i is a slack variable; C is a penalty parameter that controls the degree of penalty for misclassification; y i is the class label; is the feature vector after kernel function mapping;

[0042] State recognition and classification: Use the trained SVM classifier to classify the new feature data. Specifically, the extracted feature vector is scored and classified through the SVM model to determine which class of label it belongs to. The classification decision is made based on the relative position of the feature vector and the hyperplane, which is described by the following formula: in, To predict the class label, the state category of the sample is determined based on the calculation results of the model; w, k determine the direction and position of the classification hyperplane; is the feature vector of the current sample to be classified.

[0043] Furthermore, the anomaly detection module includes an anomaly identification unit and an anomaly reporting unit; wherein,

[0044] Abnormal identification unit: used to receive the classification results from the data analysis module and analyze the results to detect potential abnormal conditions. Specifically, the decision tree algorithm is used as the preset algorithm to evaluate the current state based on the historical performance data and known failure modes of the product. The decision tree determines whether the product is in an abnormal state through preset logical judgment. The specific algorithm is expressed as:

[0045] Among them, f1(x) and f2(x) are the feature values ​​obtained from the data analysis module, which are used for decision points in the decision tree; θ1 and θ2 are preset thresholds, which are used for classification decisions of feature values ​​in the decision tree; Class = Normal and Class = Fault are the output categories of the decision tree; Normal means that the device is in normal operation, and Fault means that the device has a fault or performance degradation;

[0046] Abnormality reporting unit: When the abnormality identification unit detects an abnormal state, the abnormality reporting unit is used to generate a detailed abnormality report, including the specific type of fault, possible causes and recommended maintenance or corrective measures. The abnormality reporting unit will ensure that all information is recorded and transmitted.

[0047] Furthermore, the adaptive learning module includes a learning interface unit, an enhanced learning unit and a parameter adjustment unit; wherein,

[0048] Learning interface unit: collects data from the signal processing module, data analysis module and anomaly detection module, including signal processing parameters, analysis results and detected anomalies, and collects data as input to the reinforcement learning algorithm to update and optimize the behavior of the model;

[0049] Reinforcement learning unit: Using the Q-learning algorithm, the signal processing and analysis parameters are dynamically adjusted according to the received data to maximize the long-term performance of the system. The operation of Q-learning is expressed by the following formula:

[0050]

[0051] Among them, Q(s, a) is the Q value of action a in state s, which represents the expected utility of the action; s t and a t are the state and action taken at time t; r t+1 Is performing action a t The reward observed later; α is the learning rate, which controls the speed of the learning process; γ is the discount factor, which represents the current value of future rewards;

[0052] Parameter adjustment unit: Based on the optimal Q value calculated by the reinforcement learning unit, it is used to adjust the parameters of the signal processing and analysis algorithms, including adjusting the filter settings, feature extraction methods, and anomaly detection thresholds to improve the accuracy and response speed of the system.

[0053] The method of the intelligent detection system of consumer electronic functional devices comprises the following steps:

[0054] S1: Collecting raw electrical signals from the consumer electronic products under test through high-sensitivity sensors;

[0055] S2: receiving the original electrical signal from step S1, and using spectrum analysis technology to identify and isolate electromagnetic interference and radio frequency interference components in the signal;

[0056] S3: receiving the signal processed in step S2, and using digital filtering and signal enhancement technology to extract and restore the interfered original signal into a pure functional signal;

[0057] S4: receiving the clean signal of step S3, extracting features through wavelet transform technology, and using support vector machine algorithm for status recognition to determine whether the functional status of the product is normal, faulty or performance degraded;

[0058] S5: receiving the analysis result of step S4, using the decision tree algorithm to determine whether there is an abnormality, and identifying the specific fault type or performance problem;

[0059] S6: Receive data from steps S3, S4 and S5, and automatically adjust parameters of signal processing and analysis algorithms through a reinforcement learning algorithm to optimize detection accuracy and system responsiveness.

[0060] Beneficial effects of the present invention:

[0061] The present invention integrates highly sensitive sensors and advanced signal processing technology, and the system can accurately capture and analyze tiny changes in electrical signals, effectively isolate and eliminate electromagnetic interference and radio frequency interference. This efficient signal processing capability ensures the high reliability of test results and reduces the misdiagnosis rate caused by environmental interference.

[0062] The present invention, by utilizing advanced data analysis technologies such as wavelet transform and support vector machine, can automatically extract key signal features and accurately identify the functional status of the product, which not only improves the level of automation of detection, but also enhances the system's ability to diagnose various abnormal conditions. Compared with traditional methods, this intelligent analysis and processing greatly improves the accuracy and response speed of fault detection, making maintenance decisions more scientific and timely.

[0063] The present invention enables the system to continuously optimize its performance according to the data collected in actual operation through the adaptive learning module. This self-optimization mechanism based on the reinforcement learning algorithm enables the system to have the ability to learn and adapt to complex electronic environments. As time goes by, the detection accuracy and reliability of the system will continue to improve, effectively supporting the needs of continuous product quality improvement and reduced maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] Figure 1 A schematic diagram of an intelligent detection system for electronic functional devices according to an embodiment of the present invention;

[0066] Figure 2 Schematic diagram of an intelligent detection method for electronic functional devices according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0068] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The words "include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0069] like Figure 1 As shown, the intelligent detection system for consumer electronic functional devices includes a signal acquisition module, an interference identification module, a signal processing module, a data analysis module, an anomaly detection module, an adaptive learning module and a central control unit; wherein,

[0070] Signal acquisition module: equipped with a high-sensitivity sensor to collect raw electrical signals from the consumer electronic products to be tested;

[0071] Interference identification module: receives the original electrical signal from the signal acquisition module, and uses the signal analysis technology of spectrum analysis to identify and isolate the electromagnetic interference and radio frequency interference components in the signal;

[0072] Signal processing module: receives the output of the interference identification module, uses digital filtering and signal enhancement technology to extract and restore the interfered original signal to a pure functional signal;

[0073] Data analysis module: receives clean signals from the signal processing module and performs advanced data processing, including feature extraction and status recognition, to determine the functional status of the product;

[0074] Anomaly detection module: receives the results of the data analysis module and uses the preset algorithm to determine whether there are any anomalies in the product function status, especially those that may be caused by interference that has not been completely eliminated; and transmits the detection results to the external communication interface for alarm or further analysis;

[0075] Adaptive learning module: By receiving data from the signal processing module, data analysis module and anomaly detection module, it uses the preset reinforcement learning algorithm to automatically adjust the signal processing parameters and analysis algorithm to optimize system performance and respond to environmental changes;

[0076] Central control unit: connects each module through the internal data bus, coordinates the data flow and processing tasks between modules, and is equipped with a processor and memory. The processor executes the program stored in the memory, manages data flow and algorithm execution, and ensures efficient and stable operation of the system.

[0077] The signal acquisition module includes a high-sensitivity sensor unit, a signal conditioning unit and a data conversion unit; wherein,

[0078] High-sensitivity sensor unit: including piezoelectric sensors and photoelectric sensors, used to collect electrical signals generated by tiny physical changes and light changes from the consumer electronic products to be tested with high precision. Piezoelectric sensors can convert mechanical pressure into electrical signals to capture tiny force changes generated during product operation; photoelectric sensors are used to detect and convert light intensity changes into electrical signals to monitor the status of light-sensitive components or display screens of products;

[0079] Signal conditioning unit: Receives the original electrical signal from the high-sensitivity sensor unit and uses a low-noise amplifier and a bandpass filter to perform preliminary signal amplification and noise filtering to optimize the readability and processability of the signal and ensure that only signals within a predetermined frequency range are transmitted to the interference identification module to reduce interference from environmental noise and non-target signals;

[0080] Data conversion unit: including analog-to-digital converter (ADC), used to convert the conditioned analog signal into a digital signal to facilitate subsequent digital processing and provide high-quality input for data analysis and processing;

[0081] Example: Assume that a motor used in a consumer electronic product mainly generates electrical signals between 1kHz and 2kHz when in operation, while common electromagnetic interference may be located in a higher frequency band, such as above 5kHz. In this case, the bandpass filter can be set to only allow frequencies in the range of 1kHz to 2kHz to pass through, which can effectively shield noise or interference of non-target frequencies and ensure that the signal transmitted to the interference identification module is clean and relevant.

[0082] In the above-mentioned signal acquisition module, through the precise acquisition of the high-sensitivity sensor unit, the effective preliminary processing of the signal conditioning unit, and the precise conversion of the data conversion unit, the entire module can effectively collect high-quality electrical signals from consumer electronic products, provide accurate input data for the system, and support subsequent interference identification and signal processing operations. This precise acquisition and processing process is the key to the system's ability to accurately perform functional detection in a high-density electronic environment.

[0083] The interference identification module includes a spectrum analysis unit, an interference detection unit and a signal isolation unit; wherein,

[0084] Spectrum analysis unit: Use fast Fourier transform (FFT) technology to perform spectrum analysis on the received original electrical signal, and convert the time domain electrical signal x(t) into the frequency domain signal X(f). The specific formula is:

[0085] Among them, x(t) represents the time domain expression of the original electrical signal, which is collected from the functional devices of consumer electronic products through the signal acquisition module; X(f) represents the converted frequency domain signal, that is, the complex representation of each frequency component obtained after the time domain signal is processed by FFT, including amplitude and phase information; f represents the frequency, the unit is Hertz (Hz), representing each frequency component in the signal; ej 2 πf t It is a complex exponential function used to convert time domain signals to frequency domain. In practical applications, the discrete form of FFT can effectively separate the amplitude of each frequency component from the continuous time series, thereby identifying the main components of the signal spectrum and the interference source.

[0086] Interference detection unit: Based on the output of the spectrum analysis unit, a preset interference frequency template is used for comparative analysis. The interference frequency template includes the frequency range of known electromagnetic interference (EMI) and radio frequency interference (RFI). The comparative analysis process formula is:

[0087]

[0088] Where D(f) represents the interference detection result. If the identified frequency component belongs to the known electromagnetic interference (EMI) or radio frequency interference (RFI) frequency range, it is marked as 1 (interference), otherwise it is 0 (no interference); X(f) is passed from the spectrum analysis unit and represents the frequency domain expression of the signal; predefined EMI / RFI ranges refer to the interference frequency ranges previously defined according to product design and environmental testing. These ranges are obtained based on historical data and frequency analysis of typical interference sources. This module identifies the frequency components that match the interference frequency template and marks them as signals that need further processing;

[0089] Signal isolation unit: Receives the interference frequency component identified by the interference detection unit, and uses a digital band-stop filter to remove the interference frequency component from the original signal. Specifically, the band-stop filter is set to suppress the interference frequency range identified in the interference detection unit. The specific operation is expressed as applying the filter function H(f), and the calculation formula is: Y(f) = X(f)·(1-H(f)), where Y(f) represents the filtered signal, that is, the frequency domain signal after the interference component is removed; X(f) is the signal transmitted from the interference detection unit, containing possible interference information; H(f) is the frequency response function of the band-stop filter, which takes a value of 1 within the interference frequency range (completely blocking the interference frequency) and a value of 0 at other frequencies (retaining the signal component).

[0090] The signal processing module includes a digital filtering unit and a signal enhancement unit; wherein,

[0091] Digital filter unit: A preset low-pass filter is used to remove high-frequency noise components and retain the low-frequency signal required for product functions. The specific cutoff frequency of the filter is set based on the operating frequency of the product and the predicted interference frequency. The low-pass filter is expressed by the following equation: Y lp (f) = X(f)·H lp (f), where Y lp (f) represents the signal after low-pass filtering, X(f) is the frequency domain signal from the interference identification module, and H lp (f) is the frequency response function of the low-pass filter, which is 1 below the cutoff frequency and rapidly approaches 0 above the cutoff frequency;

[0092] Signal enhancement unit: Use adaptive enhancement algorithm to improve the signal-to-noise ratio and strength of the signal. The enhancement algorithm analyzes the statistical characteristics of the signal, dynamically adjusts the gain coefficient, and optimizes the signal quality. The formula for signal enhancement is: Y enh (t) = g(t)·x filt (t), where Y enh (t) is the enhanced time domain signal, g(t) is the gain coefficient adaptively calculated based on the signal environment, and x filt (t) is the signal converted back to the time domain after being processed by the digital filter unit;

[0093] In the above-mentioned signal processing module, unnecessary high-frequency interference is first removed through the digital filtering unit, and then the quality and recognizability of the key signal part are improved through the signal enhancement unit. The overall operation ensures that the system can accurately restore the functional signal of the product from an environment with heavy interference. This processing method not only improves the accuracy of the signal, but also enhances the system's adaptability to complex electronic environments.

[0094] The data analysis module includes a feature extraction unit and a state recognition unit; wherein,

[0095] Feature extraction unit: used to extract features from pure signals through wavelet transform. Wavelet transform is an effective signal processing technology that can provide time and frequency information at multiple scales at the same time. It is used for feature extraction of non-stationary signals. Specifically, it decomposes the signal into wavelet coefficients to accurately capture local changes and anomalies in the signal. Such changes and anomalies will indicate potential failures in product performance indicators.

[0096] State recognition unit: Based on the wavelet transform features provided by the feature extraction unit, the support vector machine (SVM) algorithm is used for state recognition. The support vector machine is a powerful supervised learning model used for classification and regression analysis. In this unit, the support vector machine will be trained to distinguish different product functional states, including normal operation, failure mode or performance degradation. Through the above method, the system can accurately identify the functional state of consumer electronic products, thereby achieving precise monitoring of product performance and timely fault diagnosis. This advanced data processing capability greatly improves the intelligence and automation level of the system, enabling it to demonstrate efficient fault identification and state assessment functions in complex application environments.

[0097] The feature extraction unit includes:

[0098] Signal preprocessing: Before wavelet transform, the clean signal is first preprocessed, including denoising and normalization. Denoising is implemented by an adaptive filter, which automatically adjusts parameters based on the local variance of the signal to minimize the influence of residual noise; normalization ensures that the signal amplitude is within a predetermined range so that the processing result is not affected by the original strength of the signal; the preprocessing formula is:

[0099] Where x(t) is the input pure signal, which represents the electronic signal received from the signal processing module without interference and is used for further data analysis; μ x is the mean value of x(t), that is, the average level value of the signal during the entire acquisition period, obtained by taking the arithmetic mean of all the point values ​​of the signal; σ x is the standard deviation of x(t), which measures the variability or dispersion of the signal around its mean; norm (t) is a normalized signal, so that the signal has zero mean and unit variance, which is convenient for standardization and subsequent data processing steps;

[0100] Wavelet transform application: The preprocessed signal x norm (t) will be sent to the wavelet transform processor, and the Daubechies wavelet with good time-frequency localization characteristics is selected to perform multi-scale decomposition on the signal. The decomposed signal is wavelet coefficients, which are used to capture the characteristics of the signal at different scales. The specific formula is expressed as: W x (a,b)=∫x(t)ψ a，b (t)dt, where x(t) is the normalized signal after preprocessing; ψ a，b(t) is the wavelet function adjusted according to the scale parameter a and the translation parameter b. This is a mathematical function used to analyze signals at different scales and positions. a is the scale parameter, which controls the expansion and contraction of the wavelet function and affects the frequency resolution of the analysis. b is the translation parameter, which controls the movement of the wavelet function along the time axis and determines which part of the signal is analyzed. x (a, b) is the wavelet coefficient at scale a and translation parameter b, providing the frequency and time information of the signal at this scale and position;

[0101] Feature extraction: From the coefficients obtained by wavelet transform, select the coefficients that can represent the signal attributes as features, select the energy distribution that depends on the coefficients, especially those wavelet coefficients with high energy values, because they usually correspond to the important information part of the signal, and apply threshold processing technology to retain only the coefficients that exceed the threshold to eliminate the weak noise that still exists. The specific formula is:

[0102] F={W x (a i , b i )|E(W x (a i , b i ))>θ}, where W x (a i , b i ) is the coefficient obtained by wavelet transform, representing the signal at scale a i and pan b i Characteristics under E(W x (a i , b i )) is the energy of the wavelet coefficient, which is usually calculated as the square of the coefficient; θ is the energy threshold, which is used to screen those wavelet coefficients with sufficient information (i.e., high enough energy). This threshold is set based on the characteristics of the signal and the expected noise level; F is the final selected feature set, which contains all the wavelet coefficients that pass the energy threshold test. These features will be used in the subsequent state recognition steps;

[0103] The above steps not only provide a method to effectively extract key features from signals, but also ensure the reliability and representativeness of the features through improved preprocessing and feature selection steps, thereby making the subsequent state recognition more accurate and reliable.

[0104] The state recognition unit includes:

[0105] Feature preprocessing: The wavelet transform features received from the feature extraction unit are first normalized to ensure that the feature data has a uniform scale when performing SVM classification. Specifically, the features are scaled to a fixed range, such as [0, 1] or [-1, 1], so that the model training will not be biased towards features with larger values;

[0106] SVM model training: Use historical data sets to train SVM classifiers. The features of classification labels in historical data sets include normal operation, failure mode, and performance degradation. The basic principle of SVM is to find an optimal hyperplane that can maximize the interval between data points of different categories. The training of the classifier includes selecting a suitable kernel function, including a linear kernel, a polynomial kernel, or a radial basis function kernel, and adjusting the parameters of the kernel function, including the penalty parameter C and the kernel parameter p. The training process formula is expressed as:

[0107]

[0108] Among them, w is the normal vector of the hyperplane, which determines the direction of the classification hyperplane. In the classification of functional status of consumer electronic products, w helps determine how to distinguish the boundaries of different states (normal, faulty, performance degradation); k is the bias term, which together with the normal vector determines the position of the hyperplane. It adjusts the intercept of the hyperplane to optimize the classification boundary; ξ i is a slack variable that allows some data points to violate the maximum margin principle. In practical applications, this enables the SVM model to better handle noise and abnormal data. C is a penalty parameter that controls the degree of penalty for misclassification. A larger C value makes the model more severe in penalizing misclassification, which may lead to overfitting. A smaller C value makes the model more tolerant to misclassification, which may lead to underfitting. i is the class label, for each training sample, y i Indicates its true status category (for example, +1 represents normal status, 1 represents abnormal status); The feature vector after kernel function mapping is used to map the original input space to a high-dimensional feature space so that the nonlinear problem can be linearized.

[0109] State recognition and classification: Use the trained SVM classifier to classify the new feature data. Specifically, the extracted feature vector is scored and classified through the SVM model to determine which class of label it belongs to (normal, faulty, or performance degraded). The classification decision is made based on the relative position of the feature vector and the hyperplane, which is described by the following formula: in, To predict the class label, the state category of the sample is determined based on the calculation results of the model. If the result is positive, it is predicted as a positive class (normal state), and if it is negative, it is predicted as a negative class (fault or performance degradation); w, k determine the direction and position of the classification hyperplane; is the feature vector of the current sample to be classified, which is mapped by the same kernel function to adapt to the data transformation used in training the model;

[0110] Through the above steps, the state recognition unit can accurately distinguish the states of consumer electronic products between normal operation, failure mode or performance degradation, ensuring that the system can make corresponding responses and maintenance decisions based on the actual functional performance of the product. This method improves the system reliability and fault response speed and optimizes the maintenance process.

[0111] The anomaly detection module includes an anomaly identification unit and an anomaly reporting unit; wherein,

[0112] Abnormal identification unit: used to receive the classification results from the data analysis module and analyze the results to detect potential abnormal conditions. Specifically, the decision tree algorithm is used as the preset algorithm to evaluate the current state based on the historical performance data and known failure modes of the product. The decision tree determines whether the product is in an abnormal state through preset logical judgments (nodes). The specific algorithm is expressed as follows:

[0113] Among them, f1(x) and f2(x) are the feature values ​​obtained from the data analysis module, which are used for decision points in the decision tree. These feature values ​​may include features extracted from wavelet transform, outputs of support vector machine classification, or quantitative indicators generated by other mathematical transformations to express the current operating status or performance of the equipment; θ1 and θ2 are preset thresholds, which are used for classification decisions on feature values ​​in the decision tree. They are set based on historical data and statistical analysis to distinguish normal operation from potential fault states. The basis for setting thresholds may include statistical analysis, expert experience or historical fault data; Class = Normal and Class = Fault are the output categories of the decision tree; Normal means that the equipment is in normal operation, and Fault means that the equipment has a fault or performance degradation. These categories are defined based on the expected performance and safety standards of the equipment;

[0114] Abnormal Reporting Unit: When the abnormal identification unit detects an abnormal state, the abnormal reporting unit is used to generate a detailed abnormal report, including the specific type of fault, possible causes, and recommended repair or corrective measures. The abnormal reporting unit will ensure that all information is recorded and transmitted so that the maintenance team can respond quickly;

[0115] Through the above units, the anomaly detection module can effectively identify any anomalies in the product function, ensuring that the system can intervene when initial signs are detected. The use of a decision tree algorithm allows the module to make accurate judgments based on specific business rules and product performance history, while the addition of an anomaly reporting unit ensures that all relevant information can be processed in a timely manner. This system design not only improves the operational reliability of the product, but also optimizes the maintenance and service processes, reducing operational interruptions caused by failures.

[0116] The adaptive learning module includes a learning interface unit, an enhanced learning unit and a parameter adjustment unit; wherein,

[0117] Learning interface unit: collects data from the signal processing module, data analysis module and anomaly detection module, including signal processing parameters, analysis results and detected anomalies, and collects data as input to the reinforcement learning algorithm to update and optimize the behavior of the model;

[0118] Reinforcement Learning Unit: Using the Q-learning algorithm, a model-free reinforcement learning method for learning optimal strategies without full knowledge of the dynamics of the environment, the reinforcement learning unit dynamically adjusts the signal processing and analysis parameters based on the received data to maximize the long-term performance of the system. The operation of Q-learning is expressed by the following formula:

[0119]

[0120] Among them, Q(s, a) is the Q value of action a in state s, which represents the expected utility of the action; s t and a t are the state and action taken at time t; r t+1 Is performing action a t The reward observed later; α is the learning rate, which controls the speed of the learning process; γ is the discount factor, which represents the current value of future rewards;

[0121] Parameter adjustment unit: Based on the optimal Q value calculated by the reinforcement learning unit, it is used to adjust the parameters of the signal processing and analysis algorithms, including adjusting the filter settings, feature extraction methods, and anomaly detection thresholds to improve the accuracy and response speed of the system;

[0122] Through the above-mentioned units, the adaptive learning module can continuously learn and improve its performance, and optimize operating parameters based on real-time data feedback. The design of this module enables the system to not only perform well in the current environment, but also have the ability to adapt to future changes. By utilizing the reinforcement learning algorithm, the system can self-optimize without explicit guidance, thereby improving overall operational efficiency and fault handling capabilities.

[0123] like Figure 2 As shown, the method of the intelligent detection system of consumer electronic functional devices includes the following steps:

[0124] S1: Collecting raw electrical signals from the consumer electronic products under test through high-sensitivity sensors;

[0125] S2: receiving the original electrical signal from step S1, and using spectrum analysis technology to identify and isolate electromagnetic interference and radio frequency interference components in the signal;

[0126] S3: receiving the signal processed in step S2, and using digital filtering and signal enhancement technology to extract and restore the interfered original signal into a pure functional signal;

[0127] S4: receiving the clean signal of step S3, extracting features through wavelet transform technology, and using support vector machine algorithm for status recognition to determine whether the functional status of the product is normal, faulty or performance degraded;

[0128] S5: receiving the analysis result of step S4, using the decision tree algorithm to determine whether there is an abnormality, and identifying the specific fault type or performance problem;

[0129] S6: Receive data from steps S3, S4 and S5, and automatically adjust parameters of signal processing and analysis algorithms through a reinforcement learning algorithm to optimize detection accuracy and system responsiveness.

[0130] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Intelligent detection system for consumer electronic functional devices, characterized in that: It includes a signal acquisition module, an interference identification module, a signal processing module, a data analysis module, an anomaly detection module, an adaptive learning module and a central control unit; wherein, Signal acquisition module: equipped with a high-sensitivity sensor to collect raw electrical signals from the consumer electronic products to be tested; The signal acquisition module includes a high-sensitivity sensor unit, a signal adjustment unit and a data conversion unit; wherein, High-sensitivity sensor unit: including piezoelectric sensors and photoelectric sensors, used to collect electrical signals generated by tiny physical changes and light changes from the consumer electronic products to be tested with high precision. Piezoelectric sensors can convert mechanical pressure into electrical signals to capture tiny force changes generated during product operation; photoelectric sensors are used to detect and convert light intensity changes into electrical signals to monitor the status of light-sensitive components or display screens of products; Signal conditioning unit: Receives the original electrical signal from the high-sensitivity sensor unit and uses a low-noise amplifier and a bandpass filter to perform preliminary signal amplification and noise filtering to optimize the readability and processability of the signal and ensure that only signals within a predetermined frequency range are transmitted to the interference identification module to reduce interference from environmental noise and non-target signals; Data conversion unit: including an analog-to-digital converter, used to convert the conditioned analog signal into a digital signal for subsequent digital processing; Interference identification module: receives the original electrical signal from the signal acquisition module, and uses the signal analysis technology of spectrum analysis to identify and isolate the electromagnetic interference and radio frequency interference components in the signal; The interference identification module includes a spectrum analysis unit, an interference detection unit and a signal isolation unit; wherein, Spectrum analysis unit: Use fast Fourier transform technology to perform spectrum analysis on the received original electrical signal, and convert the time domain electrical signal x(t) into the frequency domain signal X(f). The specific formula is: Where x(t) represents the time domain expression of the original electrical signal; X(f) represents the converted frequency domain signal; f represents the frequency in Hertz; e -j2πft It is a complex exponential function, which is used to convert the time domain signal to the frequency domain; Interference detection unit: Based on the output of the spectrum analysis unit, a preset interference frequency template is used for comparative analysis. The interference frequency template includes the frequency range of known electromagnetic interference and radio frequency interference. The comparative analysis process formula is: Where D(f) represents the interference detection result; X(f) is transmitted from the spectrum analysis unit and represents the frequency domain expression of the signal; predefined EMI / RFI ranges refers to the interference frequency range previously defined based on product design and environmental testing; Signal isolation unit: receives the interference frequency component identified by the interference detection unit, and removes the interference frequency component from the original signal using a digital band-stop filter. Specifically, the band-stop filter is set to suppress the interference frequency range identified in the interference detection unit. The specific operation is expressed as applying the filter function H(f), and the calculation formula is: Y(f) = X(f)·(1-H(f)), where Y(f) represents the filtered signal, that is, the frequency domain signal after the interference component is removed; X(f) is the signal transmitted from the interference detection unit; H(f) is the frequency response function of the band-stop filter; Signal processing module: receives the output of the interference identification module, uses digital filtering and signal enhancement technology to extract and restore the interfered original signal to a pure functional signal; Data analysis module: receives clean signals from the signal processing module and performs advanced data processing, including feature extraction and status recognition, to determine the functional status of the product; Anomaly detection module: receives the results of the data analysis module and uses the preset algorithm to determine whether there is an anomaly in the product function status; Adaptive learning module: by receiving data from the signal processing module, data analysis module and anomaly detection module, it uses the preset reinforcement learning algorithm to automatically adjust the signal processing parameters and analysis algorithm; The adaptive learning module includes a learning interface unit, an enhanced learning unit and a parameter adjustment unit; wherein, Learning interface unit: collects data from the signal processing module, data analysis module and anomaly detection module, including signal processing parameters, analysis results and detected anomalies, and collects data as input to the reinforcement learning algorithm to update and optimize the behavior of the model; Reinforcement learning unit: Using the Q-learning algorithm, the signal processing and analysis parameters are dynamically adjusted according to the received data to maximize the long-term performance of the system. The operation of Q-learning is expressed by the following formula: Among them, Q(s, a) is the Q value of action a in state s, which represents the expected utility of the action; s t and a t are the state and action taken at time t; r t+1 Is performing action a t The reward observed later; α is the learning rate, which controls the speed of the learning process; γ is the discount factor, which represents the current value of future rewards; Parameter adjustment unit: Based on the optimal Q value calculated by the reinforcement learning unit, it is used to adjust the parameters of the signal processing and analysis algorithms, including adjusting the filter settings, feature extraction methods, and anomaly detection thresholds to improve the accuracy and response speed of the system; Central control unit: connects each module through an internal data bus, coordinates data flow and processing tasks between modules, and is configured with a processor and memory. The processor executes the program stored in the memory and manages data flow and algorithm execution.

2. The intelligent detection system for consumer electronic functional devices according to claim 1, characterized in that: The signal processing module includes a digital filtering unit and a signal enhancement unit; wherein, Digital filter unit: A preset low-pass filter is used to remove high-frequency noise components and retain the low-frequency signal required for product functions. The specific cutoff frequency of the filter is set based on the operating frequency of the product and the predicted interference frequency. The low-pass filter is expressed by the following equation: Y lp (f) = X(f)·H lp (f), where Y lp (f) represents the signal after low-pass filtering, X(f) is the frequency domain signal from the interference identification module, and H lp (f) is the frequency response function of the low-pass filter, which is 1 below the cutoff frequency and rapidly approaches 0 above the cutoff frequency; Signal enhancement unit: Use adaptive enhancement algorithm to improve the signal-to-noise ratio and strength of the signal. The enhancement algorithm analyzes the statistical characteristics of the signal, dynamically adjusts the gain coefficient, and optimizes the signal quality. The formula for signal enhancement is: Y enh (t) = g(t)·x filt (t), where Y enh (t) is the enhanced time domain signal, g(t) is the gain coefficient adaptively calculated based on the signal environment, and x filt (t) is the signal converted back to the time domain after being processed by the digital filter unit.

3. The intelligent detection system for consumer electronic functional devices according to claim 1, characterized in that: The data analysis module includes a feature extraction unit and a state recognition unit; wherein, Feature extraction unit: used to extract features from pure signals through wavelet transform. Wavelet transform can provide time and frequency information at multiple scales at the same time, and is used for feature extraction of non-stationary signals. Specifically, the signal is decomposed into wavelet coefficients to accurately capture local changes and anomalies in the signal, which will indicate potential failures in product performance indicators. State recognition unit: Based on the wavelet transform features provided by the feature extraction unit, the support vector machine (SVM) algorithm is applied for state recognition. The support vector machine will be trained to distinguish different product functional states, including normal operation, failure mode or performance degradation.

4. The intelligent detection system for consumer electronic functional devices according to claim 3, characterized in that: The feature extraction unit comprises: Signal preprocessing: Before wavelet transform, the clean signal is first preprocessed, including denoising and normalization. Denoising is implemented by an adaptive filter, which automatically adjusts parameters based on the local variance of the signal to minimize the influence of residual noise; normalization ensures that the signal amplitude is within a predetermined range so that the processing result is not affected by the original strength of the signal; the preprocessing formula is: Among them, x(t) is the pure input signal; μ x is the mean value of x(t), that is, the average level of the signal during the entire acquisition period; σ x is the standard deviation of x(t); norm (t) is the normalized signal; Wavelet transform application: The preprocessed signal x norm (t) will be sent to the wavelet transform processor, and the Daubechies wavelet with good time-frequency localization characteristics is selected to perform multi-scale decomposition on the signal. The decomposed signal is wavelet coefficients, which are used to capture the characteristics of the signal at different scales. The specific formula is expressed as: W x (a, b) = ∫x(t)ψ a，b (t)dt, where x(t) is the normalized signal after preprocessing; ψ a，b (t) is the wavelet function adjusted according to the scale parameter a and the translation parameter b; a is the scale parameter; b is the translation parameter; W x (a, b) is the wavelet coefficient at scale a and translation parameter b, providing the frequency and time information of the signal at this scale and position; Feature extraction: From the coefficients obtained by wavelet transform, select the coefficients that can represent the signal attributes as features, and apply threshold processing technology to retain only the coefficients exceeding the threshold to eliminate the weak noise that still exists. The specific formula is: F={W x (a i , b i )|E(W x (a i , b i ))>θ}, where W x (a i , b i ) is the coefficient obtained by wavelet transform, representing the signal at scale a i and pan b i Characteristics under E(W x (a i , b i )) is the energy of the wavelet coefficient; θ is the energy threshold; F is the final selected feature set.

5. The intelligent detection system for consumer electronic functional devices according to claim 4, characterized in that: The state recognition unit comprises: Feature preprocessing: The wavelet transform features received from the feature extraction unit are first normalized to ensure that the feature data has a uniform scale when performing SVM classification; SVM model training: Use historical data sets to train SVM classifiers. The features of classification labels in the historical data sets include normal operation, failure mode, and performance degradation. Classifier training includes selecting appropriate kernel functions, including linear kernels, polynomial kernels, or radial basis function kernels, and adjusting the parameters of the kernel function, including the penalty parameter C and the kernel parameter p. The training process formula is expressed as: Among them, w is the normal vector of the hyperplane, which determines the direction of the classification hyperplane; k is the bias term, which together with the normal vector determines the position of the hyperplane; ξ i is a slack variable; C is a penalty parameter that controls the degree of penalty for misclassification; y i is the class label; is the feature vector after kernel function mapping; State recognition and classification: Use the trained SVM classifier to classify the new feature data. Specifically, the extracted feature vector is scored and classified through the SVM model to determine which class of label it belongs to. The classification decision is made based on the relative position of the feature vector and the hyperplane, which is described by the following formula: in, To predict the class label, the state category of the sample is determined based on the calculation results of the model; w, k determine the direction and position of the classification hyperplane; is the feature vector of the current sample to be classified.

6. The intelligent detection system for consumer electronic functional devices according to claim 1, characterized in that: The anomaly detection module includes an anomaly identification unit and an anomaly reporting unit; wherein, Abnormal identification unit: used to receive the classification results from the data analysis module and analyze the results to detect potential abnormal conditions. Specifically, the decision tree algorithm is used as the preset algorithm to evaluate the current state based on the historical performance data and known failure modes of the product. The decision tree determines whether the product is in an abnormal state through preset logical judgment. The specific algorithm is expressed as: Among them, f1(x) and f2(x) are the feature values ​​obtained from the data analysis module, which are used for decision points in the decision tree; θ1 and θ2 are preset thresholds, which are used for classification decisions of feature values ​​in the decision tree; Class = Normal and Class = Fault are the output categories of the decision tree; Normal means that the device is in normal operation, and Fault means that the device has a fault or performance degradation; Abnormality reporting unit: When the abnormality identification unit detects an abnormal state, the abnormality reporting unit is used to generate a detailed abnormality report, including the specific type of fault, possible causes and recommended maintenance or corrective measures. The abnormality reporting unit will ensure that all information is recorded and transmitted.

7. The method of the intelligent detection system of consumer electronic functional devices according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1: Collecting raw electrical signals from the consumer electronic products under test through high-sensitivity sensors; S2: receiving the original electrical signal from step S1, and using spectrum analysis technology to identify and isolate electromagnetic interference and radio frequency interference components in the signal; S3: receiving the signal processed in step S2, and using digital filtering and signal enhancement technology to extract and restore the interfered original signal into a pure functional signal; S4: receiving the clean signal of step S3, extracting features through wavelet transform technology, and using support vector machine algorithm for status recognition to determine whether the functional status of the product is normal, faulty or performance degraded; S5: receiving the analysis result of step S4, using the decision tree algorithm to determine whether there is an abnormality, and identifying the specific fault type or performance problem; S6: Receive data from steps S3, S4 and S5, and automatically adjust parameters of signal processing and analysis algorithms through a reinforcement learning algorithm to optimize detection accuracy and system responsiveness.

Images