Handheld Electronic Component-Based Inspection and Analysis Method, System, and Device

Abstract

The present invention relates to the technical field of electronic component detection, and discloses a handheld electronic component inspection and analysis method, system and device, including: emitting a fundamental wave in the S band to a target object; acquiring a time data sequence signal received by a receiving end; implementing a multi-convolution feature extraction strategy to extract fundamental wave features, second harmonic features and third harmonic features in the time data sequence signal, and obtaining a feature map of the time step sequence; according to the feature map of the time step sequence, implementing a timing dependence capture strategy to capture the dependence relationship of the time data sequence signal; implementing an adaptive sensitivity adjustment strategy to automatically adjust the sensitivity of a detector based on environmental noise and detection distance; implementing a Fourier frequency domain analysis strategy on the time data sequence signal to predict the time data sequence signal at a future moment; if environmental noise affects the signal, implementing an environmental filtering adaptive noise reduction strategy to reduce the interference of environmental noise and enhance the signal recognition ability.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic component detection, and specifically to a handheld method, system and device for inspecting and analyzing electronic components. Background Art

[0002] The handheld communication product detector has high detection accuracy, simple operation, light and beautiful appearance, clear and loud sound, and has two alarm modes of sound and vibration. It is especially suitable for use in secret-related places with extremely high security inspection requirements, such as examination rooms, airports, public security, etc. Electronic components are the core components of handheld communication product detectors, and their performance directly determines the overall reliability and stability of electronic devices. With the gradual expansion of the application scope of electronic products, especially in high-end fields such as aerospace, automotive, communication, and medical, higher requirements are put forward for the quality inspection of electronic components.

[0003] In the field of modern electronic component detection, traditional detection methods usually rely on manual inspection and simple electronic test equipment, and it is difficult to accurately identify tiny defects or complex signal characteristics. In the face of high-density integrated circuits or complex electronic systems, missed inspections or misinspections are likely to occur. Manual detection requires a large amount of time and manpower, especially in a large-scale production environment, and the efficiency problem is particularly prominent. The degree of automation is low, and it cannot quickly adapt to the detection requirements of different types of components. Traditional detection equipment is easily interfered in a noisy environment, resulting in unstable detection results. There is a lack of effective noise filtering and sensitivity adjustment mechanisms, making the performance of the equipment inconsistent in different environments. Usually, there is a lack of the ability to deeply analyze detection data, and valuable information cannot be extracted from complex signals. Most detection equipment lacks intelligent feature extraction and analysis capabilities and cannot perform adaptive adjustment and optimization based on detection results. This limits the application of the equipment in a dynamically changing environment.

[0004] This solution proposes a handheld method, system and device for inspecting and analyzing electronic components, which realizes multi-band analysis of the component state by transmitting an S-band fundamental wave signal and capturing its harmonic reflection signal, and improves the accuracy of monitoring electronic components. Summary of the Invention

[0005] The present invention provides a handheld method, system and device for inspecting and analyzing electronic components, which helps to solve the problems mentioned in the above background art.

[0006] In a first aspect, the present application provides a handheld method for inspecting and analyzing electronic components, and adopts the following technical solution: A handheld method for inspecting and analyzing electronic components includes: emitting a fundamental wave of the S-band from a transmitting end to a target object;

[0007] The fundamental wave reflected by the target object and the second harmonic and third harmonic generated by the target object are captured by the receiving end;

[0008] Obtain the time data sequence signal x(t) = {x1, x2,..., x t}, where x i represents the discretized time data sequence signal, t is the time step, and t is a positive integer;

[0009] Execute a multi-convolution feature extraction strategy to extract the fundamental wave feature z1, second harmonic feature z2, and third harmonic feature z3 from the time data sequence signal, and obtain the feature map Z = {z1, z2, z3} of the time step sequence;

[0010] According to the feature map of the time step sequence, execute a temporal dependence capture strategy to capture the dependence relationship of the time data sequence signal; obtain the environmental noise σ(t) and detection distance d(t) at time t;

[0011] Execute an adaptive sensitivity adjustment strategy to automatically adjust the sensitivity of the detector based on the environmental noise and detection distance;

[0012] Execute a Fourier frequency domain analysis strategy on the time data sequence signal to predict the time data sequence signal at a future time, and based on the predicted time data sequence signal, determine whether the environmental noise affects the signal;

[0013] If the environmental noise affects the signal, execute an environmental filtering adaptive noise reduction strategy to reduce the interference of the environmental noise.

[0014] Through this handheld electronic component inspection and analysis method, accurate target detection and signal processing can be achieved. First, the S-band fundamental wave emitted by the transmitting end to the target object, combined with the reflected signal captured by the receiving end and the second harmonic and third harmonic generated by the target object, can comprehensively capture the different frequency information of the target object. Through the multi-convolution feature extraction strategy of extracting the fundamental wave, second harmonic, and third harmonic features, key features with discriminability can be effectively extracted from the signal. This way of feature extraction makes the signal processing process more efficient and accurate, avoiding redundancy and inefficiency in traditional methods. Combining the adaptive sensitivity adjustment strategy of environmental noise and detection distance, the system can dynamically adjust the sensitivity of the detector, thus maintaining a high detection accuracy under different environmental conditions. This automatic adjustment of sensitivity can cope with the interference of environmental noise and maintain stable performance during long-term and long-distance detection. Finally, through the Fourier frequency domain analysis strategy and adaptive noise reduction processing, the system can effectively filter out the interference of environmental noise on the signal, thus ensuring the accuracy and reliability of the data.

[0015] Preferably, the execution of the multi-convolution feature extraction strategy for extracting the fundamental wave, second harmonic, and third harmonic in the time data sequence signal includes:

[0016] Perform a convolution operation:

[0017] Set the frequency of the fundamental wave as f1;

[0018] Then the frequency of the second harmonic is f2 = 2f1, and the frequency of the third harmonic is f3 = 3f1;

[0019] Perform a convolution operation on the fundamental wave to obtain a fundamental wave feature map

[0020] Perform a convolution operation on the second harmonic to obtain a second harmonic feature map

[0021] Perform a convolution operation on the third harmonic to obtain a third harmonic feature map Where, x i+j-1 is the discrete value of the time data sequence signal at the i + j - 1 moment, w 1j , w 2j , w 3j are respectively the j-th weights of the convolution kernels of the fundamental wave, second harmonic, and third harmonic, and b1, b2, b3 are respectively the bias terms of the fundamental wave, second harmonic, and third harmonic;

[0022] Perform a pooling operation:

[0023] Max pooling operation Where, i = 1, 2, 3, z i is the output of the convolutional layer, pool_region is the region of the pooling operation window, and p i is the output result after pooling.

[0024] The introduction of convolution operations and pooling operations helps to further extract the features of the target signal. By performing convolution operations on the fundamental wave, second harmonic, and third harmonic, the system can carefully analyze the characteristics of different frequency components in the frequency domain. The feature maps extracted by the convolution operation provide a multi-dimensional signal representation, enabling subsequent analysis to more precisely capture the key information in the signal. In addition, through the max pooling operation, the system can effectively reduce the dimension of the feature map, reduce the computational amount, while retaining the most significant features, thereby improving the efficiency of the system when processing large-scale data. This operation method optimizes the use of computing resources while ensuring the effectiveness of the features, making the entire processing process more efficient. Through this multi-convolution and pooling method, the system can more flexibly handle different types of signals, especially in the face of complex and high-noise environments, it can more accurately extract useful signals, improving the overall performance and stability of the system.

[0025] Preferably, for the feature map according to the time step sequence, a temporal dependence capture strategy is executed to capture the dependence relationship of the time data sequence signal, including:

[0026] Execute an input gate, and the input gate controls the current input z t and the hidden state h of the previous time step t-1 to determine how to update the content i of the memory cell t i = σ × (W i × [h t-1 , z t + b i ), where i t is the activation value of the input gate, which determines which parts of the current input will be written into the memory cell; σ is the sigmoid activation function, and the output range is between [0, 1], W i is the weight matrix of the input gate, and b i is the bias term of the input gate;

[0027] Execute a forget gate, f t f = σ × (W f × [h t-1 , z t + b f ), where f t is the activation value of the forget gate, which controls how much of the content stored in the memory cell is forgotten, W f is the weight matrix of the forget gate, and b f is the bias term of the forget gate.

[0028] By introducing the LSTM (Long Short-Term Memory) model, the temporal modeling ability of signal processing is further improved. LSTM can capture the long-term dependence relationship in the time series signal, which is crucial for analyzing complex signals. The design of the input gate, forget gate, and output gate enables the system to flexibly control the flow of information and the update of memory. By controlling which information enters the memory cell through the input gate and which information is forgotten through the forget gate, the system can maintain the transmission of key information in the signal processing of long time series, while suppressing the interference of irrelevant information. This gating mechanism allows LSTM to perform well in processing complex data. Especially when facing time series data with long-term dependence relationships, it can effectively capture the long-term effects, improving the accuracy and reliability of prediction. The output gate ensures a tight association between the final output result and the memory state, ensuring the interpretability of the output and the transparency of the system. Through such temporal modeling, the system can more accurately analyze and predict the dynamic behavior of the target object, thereby improving the accuracy of target recognition and analysis. The introduction of LSTM enables the system to exhibit stronger robustness and adaptability when facing high-dimensional, non-linear, and long-term dependent time series signals.

[0029] Preferably, for the feature map according to the time step sequence, a temporal dependence capture strategy is executed to capture the dependence relationship of the time data sequence signal, including:

[0030] Update the memory cell, c t = f t × c t-1 + i t × tanh(W c [h t-1, z t + b c ), where f t × c t-1 represents the information retained from the previous moment's memory determined by the forget gate, and i t × tanh(W c [h t-1, z t + b c ) is the update of the memory cell under the control of the input gate by the current input z t . tanh is the hyperbolic tangent activation function, ensuring that the new memory content is restricted within an appropriate range;

[0031] Execute the output gate, o t = σ × (W o × [h t-1 , z t + b o ), where o t is the activation value of the output gate, determining which information in the memory cell c t will affect the final output. W o is the weight matrix of the output gate, and b o is the bias term of the output gate;

[0032] Hidden state, the output h t of the LSTM is obtained by controlling the influence of the memory cell c t through the output gate and calculating it through the tanh activation function. The formula is h t = o t × tanh(c t );

[0033] Obtain the LSTM output function y = softmax(W f × h t + b f ), where W f is the weight matrix of the forget gate, and b f is the bias term of the forget gate.

[0034] Through the update mechanism of the memory unit and the design of the output gate, it is ensured that historical information can be fully utilized at each time step, and the transmission of information flow can be effectively controlled. The collaborative effect of the forget gate and the input gate enables the system to make flexible adjustments for different situations. For example, in the case of high noise or low information volume, the system can improve efficiency by forgetting unnecessary historical information and retaining important information. By controlling the update range of information through the hyperbolic tangent activation function (tanh), the system can avoid problems such as information explosion or gradient disappearance, ensuring the stability of the model during long-term training. In addition, the output gate controls the final output result, ensuring that the LSTM model can accurately output the analysis and prediction results of the target object based on the current state. Generally speaking, the design of this section further improves the adaptability of the LSTM network in a dynamic environment, making the system more accurate and stable when processing complex time-series data.

[0035] Preferably, the implementation of the adaptive sensitivity adjustment strategy automatically adjusts the sensitivity of the detector based on environmental noise and detection distance, including:

[0036] Calculating the sensitivity where S(t) represents the sensitivity at time t, S0 represents the initial sensitivity, and α represents the adjustment coefficient;

[0037] Correspondingly, when the environmental noise σ(t) increases, the sensitivity of the detector increases;

[0038] When the detection distance d(t) increases, the sensitivity of the detector increases.

[0039] The adaptive adjustment strategy of sensitivity ensures that the detector can automatically adjust its sensitivity according to the noise level and detection distance of different environments. This dynamic adjustment mechanism enables the system to maintain a relatively stable and accurate detection effect under different working conditions. When the environmental noise increases, the system automatically increases the sensitivity to capture weak signals; when the detection distance increases, the system also adjusts the sensitivity to expand the detection range. The introduction of this adaptive mechanism not only improves the performance of the system in complex environments but also ensures that the detector can operate continuously and stably for a long time, effectively avoiding the limitations of fixed sensitivity settings. The adjustment of sensitivity is a real-time and automated process that can cope with various sudden environmental changes and maintain the optimization of detection performance.

[0040] Preferably, the implementation of the Fourier frequency-domain analysis strategy for the time data sequence signal predicts the time data sequence signal at a future time, and based on the predicted time data sequence signal, determines whether the environmental noise affects the signal, including:

[0041] Obtaining the time data sequence signal x(t);

[0042] Use the Fourier transform to convert the signal from the time domain to the frequency domain where \(X(f)\) is the frequency-domain signal, \(x(t)\) is the time-domain signal, and \(f\) is the frequency;

[0043] Set the frequency interval;

[0044] Determine whether the frequency-domain signal obtained by using the Fourier transform is within the frequency interval. If it exists, it is determined that the environmental noise does not affect the signal;

[0045] If it does not exist, it is determined that the environmental noise affects the signal;

[0046] When the environmental noise affects the signal, execute the environmental filtering adaptive noise reduction strategy to reduce the interference of the environmental noise. Specifically:

[0047] Add the time data sequence signal and the environmental noise to the filter. Then the size of the signal received by the filter is \(y(t)=x(t)+\sigma(t)\); the size of the signal output by the filter is where \(*\) is the convolution operation and \(w(t)\) is the weight vector of the adaptive filter;

[0048] Calculate the filtering noise reduction error Update the filter weight using the minimum filtering mean square error. Specifically:

[0049] Calculate the average value of the square of the filtering noise reduction error:

[0050] where \(E(\cdot)\) represents the expected value of the filtering noise reduction error;

[0051] Update the weight vector \(w(t + 1)=w(t)+\mu\times e(t)\times y(t)\), where \(\mu\) is the learning rate that controls the step size of the weight update, and \(e(t)\) is the current error;

[0052] Use the updated weight vector to calculate the new filtering noise reduction error and update the average value of the square of the filtering noise reduction error;

[0053] Set the error threshold, compare the average value of the square of the filtering noise reduction error with the error threshold. If the filtering noise reduction error is less than or equal to the error threshold, stop updating the filter weight.

[0054] By introducing the Fourier frequency-domain analysis strategy, the system can determine whether the signal is affected by noise in the frequency domain. This frequency-domain analysis can not only reduce the computational complexity in the time domain but also enable the system to more sensitively identify the interference of noise on the signal. Especially when the environmental noise in the high-frequency band is significant, it can effectively distinguish the influence of noise. Through the adaptive filtering noise reduction strategy, the system can make flexible adjustments according to the type of noise, thereby minimizing interference under different noise conditions. The weight update mechanism of the filter and the least mean square algorithm ensure that the filtering process can continuously optimize the noise suppression effect and guarantee the accuracy and reliability of the signal. This strategy has great advantages in dealing with complex signals and high-noise environments, effectively improving the robustness and sensitivity of the system.

[0055] In a second aspect, the present application provides a handheld electronic component inspection and analysis system applied to the above-mentioned handheld electronic component inspection and analysis method, and adopts the following technical solutions: A handheld electronic component inspection and analysis system includes:

[0056] Signal acquisition and multi-band analysis module: Perform multi-band analysis through the fundamental wave signal in the S band and its harmonic reflection signals;

[0057] Convolution feature extraction module: Extract the fundamental wave and harmonic features in the signal through convolution and pooling;

[0058] Temporal dependence capture module: The LSTM network captures the temporal dependence in the signal for long-term signal prediction;

[0059] Sensitivity adjustment module: Adaptively adjust the detector sensitivity to adapt to different environmental noises and detection distances;

[0060] Fourier analysis and noise reduction module: Perform frequency-domain analysis through Fourier transform and combine with adaptive filtering to reduce environmental noise interference.

[0061] In a third aspect, the present application provides a handheld electronic component inspection and analysis device applied to the above-mentioned handheld electronic component inspection and analysis method, and adopts the following technical solutions: A handheld electronic component inspection and analysis device includes transmitting signals using a high-speed data bus, the radio frequency transmitting the fundamental wave signal, the receiving module receiving the signal reflected by the target object, converting the received signal through analog-to-digital conversion and then transmitting it to the digital signal processing unit, the digital signal processing unit processing the signal and then transmitting it to the deep learning acceleration unit for feature extraction and temporal analysis, the signal output by the deep learning unit is used in cooperation with the Fourier analysis and noise reduction module to judge the influence of environmental noise, the environmental monitoring and sensing device collects environmental data and then transmits it to the sensitivity adjustment module to adjust the detector sensitivity, and interacts with the user through the user interface.

[0062] The present invention has the following beneficial effects:

[0063] 1. This handheld inspection and analysis method based on electronic components can significantly enhance the signal recognition ability of target objects by capturing the fundamental wave, reflected signal, second harmonic, and third harmonic in multiple frequency bands at the transmitting end. The transmitted fundamental wave in the S band and the reflected signal at the receiving end, combined with the second and third harmonics generated by the target object, not only improve the multi-dimensional information collection of the signal but also enhance the recognition ability of complex target objects. Through the multi-convolution feature extraction strategy, the system can extract the feature maps of the fundamental wave, second harmonic, and third harmonic from the time data sequence, which provides a more accurate basis for subsequent signal analysis. The introduction of multiple convolutional layers can effectively analyze the signal at different levels, extract more significant and accurate features, especially in the face of complex or noisy environments, and accurately distinguish the frequency band characteristics of the signal. Under the influence of environmental noise and detection distance, the adaptive adjustment strategy of the detector sensitivity can adjust the sensitivity in real time according to external conditions, thus ensuring the detection accuracy and stability. The system can also perform frequency domain analysis on the received signal through the Fourier frequency domain analysis strategy, accurately predict the signal changes at future moments, and effectively judge whether environmental noise interferes with the signal.

[0064] 2. This handheld inspection and analysis method based on electronic components uses convolution operations to extract the feature maps of the fundamental wave, second harmonic, and third harmonic. Through convolution operations on different frequency components, the system can deeply analyze the frequency characteristics of the signal and reveal the mutual relationship between different harmonics. The introduction of convolution operations enables the system to accurately extract features from the signal through specific convolutional kernels, and the weights and bias terms of each convolutional kernel can be dynamically adjusted according to the actual situation of the signal, thus ensuring the flexibility and accuracy of feature extraction. At the same time, the addition of pooling operations, especially the max pooling operation, can effectively reduce the data dimension, thereby reducing the computational amount and optimizing the computational efficiency. The essence of the pooling operation is to reduce the dimension of the feature map. It retains the most significant features and discards redundant information, not only improving the computational efficiency of the system but also making the subsequent analysis process more efficient. This operation can further improve the performance of the system when processing large-scale data and ensure that the feature extraction process is not interfered by redundant information. Overall, through the combination of convolution operations and pooling operations, the system can more efficiently extract and optimize signal features, improve the accuracy and computational efficiency of feature analysis, especially in the face of complex and noisy environments, and can more stably and efficiently extract key information, further improving the performance of the system.

[0065] 3. This handheld electronic component inspection and analysis method can capture long-term dependencies in time series signals by introducing the LSTM (Long Short-Term Memory) model, further enhancing the ability to process signal dynamic changes. The design of LSTM controls the flow of information through input gates, forget gates, and output gates, ensuring the effective transmission and update of signals between different time steps. The introduction of input gates can control which parts of the current input signal enter the memory unit, avoiding interference from redundant information to the memory unit. Through the cooperation of forget gates and input gates, the system can process changes in time series signals more precisely, making the capture of long-term dependencies more reliable and stable. The output gate in LSTM determines the information transmitted from the memory unit to the output layer, which regulates the output content according to the current hidden state, ensuring that the final output result can accurately reflect the dynamic changes of the signal. By introducing the LSTM model, the system can process complex time series signals more efficiently. Especially when there are long time intervals of dependencies in the signal, LSTM can effectively capture these long-term information, improving the accuracy and reliability of prediction.

[0066] 4. This handheld electronic component inspection and analysis method further enhances the ability to process time series signals by refining the working mechanism of LSTM. The update mechanism of the memory unit and the design of the output gate in LSTM can ensure that at each time step, the system can make full use of historical information and flexibly adjust according to different inputs. The design of forget gates and input gates ensures that the system can dynamically adjust the memory content according to the characteristics of the signal, guaranteeing the accurate analysis of the signal. In the process of updating the memory unit, the use of the tanh activation function can ensure that the new memory content is updated within a reasonable range, avoiding excessive or too small update amplitudes, thus improving the stability and accuracy of the model. Through the collaborative action of forget gates and input gates in LSTM, the system can flexibly adjust the memory state and effectively reduce the interference of irrelevant information.

[0067] 5. The handheld electronic component inspection and analysis method. The adaptive sensitivity adjustment strategy enables the detector to dynamically adjust the sensitivity according to different environmental noises and detection distances, thus ensuring that the system always maintains the best performance under different working conditions. By calculating and adjusting the sensitivity in real time, the system can increase the sensitivity when the environmental noise is large, so as to ensure that weak signals can be captured; when the detection distance is far, the system can expand the sensitivity and extend the detection range. This adaptive adjustment mechanism not only optimizes the performance of the detector in complex environments, but also can cope with external environmental changes, avoiding the limitations brought by fixed sensitivity settings. The automatic adjustment of sensitivity can ensure that the detector can play its maximum role in environments with high noise or weak signals, improving the accuracy and stability of the signals. Overall, this strategy can dynamically optimize the system performance according to the different characteristics of the external environment, enhancing the adaptability and reliability of the system under different working conditions.

[0068] 6. The handheld electronic component inspection and analysis method. Through the Fourier frequency domain analysis strategy, the system can accurately analyze the spectral characteristics of the signal, thereby judging whether the environmental noise interferes with the signal. The Fourier transform can convert the time-domain signal into a frequency-domain signal, avoiding the complexity in time-domain analysis. Especially when the signal has multiple frequency components, frequency-domain analysis can more intuitively reveal the influence of noise. By setting the frequency interval, the system can judge whether the frequency-domain signal is within the frequency band affected by noise and make corresponding processing based on this information. When the environmental noise affects the signal, the system will perform noise reduction processing through the adaptive filtering strategy to reduce the interference of noise on the signal. The weight update of the filter and the least mean square algorithm ensure the continuous optimization of the noise reduction effect, enhancing the adaptability of the system to noise changes. Through these frequency-domain analysis and noise reduction processing strategies, the system can effectively improve the accuracy of the signal, avoid the interference caused by noise, and thus ensure the accuracy and reliability of the signal processing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a schematic flow chart of the method of the present invention.

[0070] Figure 2 It is a schematic structural diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0071] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0072] Example 1. Refer toFigure 1 , a handheld inspection and analysis method based on electronic components, comprising:

[0073] Sending a fundamental wave in the S band from the transmitting end to the target object;

[0074] Capturing the fundamental wave reflected by the target object and the second and third harmonics generated by the target object by the receiving end;

[0075] Obtaining the time data sequence signal x(t) = {x1, x2, …, x t}, where x i represents the discretized time data sequence signal, t is the time step number, and t is a positive integer;

[0076] Executing a multi-convolution feature extraction strategy to extract the fundamental wave feature z1, second harmonic feature z2, and third harmonic feature z3 in the time data sequence signal, and obtaining the feature map Z = {z1, z2, z3} of the time step sequence;

[0077] According to the feature map of the time step sequence, executing a temporal dependence capture strategy to capture the dependence relationship of the time data sequence signal; obtaining the environmental noise σ(t) and detection distance d(t) at time t;

[0078] Executing an adaptive sensitivity adjustment strategy to automatically adjust the sensitivity of the detector based on the environmental noise and detection distance;

[0079] Performing a Fourier frequency domain analysis strategy on the time data sequence signal to predict the time data sequence signal at a future moment, and judging whether the environmental noise affects the signal based on the predicted time data sequence signal;

[0080] If the environmental noise affects the signal, then execute an environmental filtering adaptive noise reduction strategy to reduce the interference of the environmental noise.

[0081] By introducing a capture and feature extraction strategy for multi-band signals, the object recognition ability of the system is effectively enhanced. The capture of the fundamental wave signal at the S-band transmitter and receiver for the reflected signal of the target object, combined with the extraction of the second and third harmonics, can provide the system with richer target information. This signal capture based on frequency domain information and multi-band signal analysis enables the system to obtain more dimensional information when facing target objects. Especially in complex or highly interfering environments, it can significantly improve the accuracy of signal recognition and analysis. Through the multi-convolution feature extraction strategy, the system can accurately extract the features of the fundamental wave, second harmonic, and third harmonic from the time data sequence, effectively extracting the features of the signal using convolutional kernels, and further enhancing the signal identification ability. The convolution operation can not only effectively improve the accuracy of signal feature extraction, but also through the setting of multiple convolutional layers, the system can gradually extract more refined features from different levels, making the system more robust and reliable in complex environments. The temporal dependence capture strategy further enhances the prediction ability and adaptability to dynamic target objects by capturing the long-term dependence relationship of signals in the time series. The adaptive sensitivity adjustment mechanism for environmental noise and detection distance ensures that the detector can maintain stable detection accuracy under different environmental conditions by automatically adjusting the detector sensitivity.

[0082] Execute the multi-convolution feature extraction strategy to extract the fundamental wave, second harmonic, and third harmonic from the time data sequence signal, including:

[0083] Perform convolution operations:

[0084] Set the frequency of the fundamental wave to f1;

[0085] Then the frequency of the second harmonic is f2 = 2f1, and the frequency of the third harmonic is f3 = 3f1;

[0086] Perform a convolution operation on the fundamental wave to obtain the fundamental wave feature map

[0087] Perform a convolution operation on the second harmonic to obtain the second harmonic feature map

[0088] Perform a convolution operation on the third harmonic to obtain the third harmonic feature map where x i+j-1 is the discrete value of the time data sequence signal at the i + j - 1 moment, w 1j , w 2j , w 3j are respectively the j-th weights of the convolutional kernels of the fundamental wave, second harmonic, and third harmonic, and b1, b2, b3 are respectively the bias terms of the fundamental wave, second harmonic, and third harmonic;

[0089] Perform pooling operations:

[0090] Max pooling operation where \(i = 1, 2, 3\), \(z\) i is the output of the convolutional layer, pool_region is the region of the pooling operation window, \(p\) i is the output result after pooling.

[0091] After setting the frequencies of the fundamental wave, second harmonic, and third harmonic, the convolution operation can effectively extract features for signals of different frequency components, helping the system accurately identify the signal features in different frequency bands. This process depends on the dynamic weight adjustment of the convolution kernel. By continuously optimizing the parameters of the convolution kernel, the system can flexibly respond in different signal environments and achieve accurate feature extraction. Based on the convolution operation, the max pooling operation effectively reduces the dimension of the feature map, thereby reducing the computational amount and resource consumption, while retaining important feature information. The pooling operation can not only reduce the impact of redundant information on the system performance but also improve the computational efficiency of the system, enabling the system to maintain high performance and response speed when processing large-scale data. Through the combination of convolution and pooling operations, the system can extract the key information of the signal in a more efficient manner, thus ensuring the accuracy of subsequent analysis and prediction. It maintains high precision and stability, especially in the face of a changing and highly interfering signal environment, can accurately extract effective signal features, and avoid the influence of information loss or interference on the results.

[0092] According to the feature map of the time step sequence, execute the temporal dependence capture strategy to capture the dependence relationship of the time data sequence signal, including:

[0093] Execute the input gate, and the input gate controls how the current input \(z\) t and the hidden state \(h\) of the previous time step t-1 update the content \(i\) of the memory unit t \(= \sigma\times(W\) i \times[h\) t-1 , z\) t +b\) i ), where \(i\) t is the activation value of the input gate, which determines which parts of the current input will be written into the memory unit; \(\sigma\) is the sigmoid activation function, and the output range is between \([0, 1]\), \(W\) i is the weight matrix of the input gate, \(b\) i is the bias term of the input gate;

[0094] Execute the forget gate, \(f\) t \(= \sigma\times(W\) f \times[h\) t-1 , z\) t +b\) f ), where \(f\) tis the activation value of the forget gate, controlling how much of the content stored in the memory cell is forgotten, W f is the weight matrix of the forget gate, b f is the bias term of the forget gate.

[0095] By introducing the LSTM (Long Short-Term Memory network), the processing ability of the system for time series data is improved. LSTM is a neural network model that can capture long-term dependencies and shows excellent performance when processing time-series signals. Through the design of the input gate, forget gate and output gate, LSTM can effectively process the signals at each time step, ensuring that the system can dynamically update and adjust the memory cell according to historical information. The input gate controls which input signals are written into the memory cell, and the forget gate determines which content needs to be forgotten based on historical information. The cooperation of these two gates enables the system to maintain efficient information flow between different time steps, thus effectively processing long-term dependence signals. The output gate determines the content extracted from the memory cell, ensuring that the system can output accurate predictions of the signal according to the current state. The introduction of the LSTM model improves the prediction ability of the system for dynamic signals, especially when processing signals with long-term dependence relationships, and can provide more accurate and reliable results. The design of the LSTM model not only improves the prediction accuracy of the system, but also enhances the robustness of the system in processing complex time-series signals, enabling the system to maintain high stability when facing a dynamically changing environment.

[0096] According to the feature map of the time step sequence, execute the time series dependence capture strategy to capture the dependence relationship of the time data series signal, including:

[0097] Update the memory cell, c t = f t × c t-1 + i t × tanh(W c [h t-1, z t + b c ), where, f t × c t-1 represents the information retained from the previous moment's memory determined by the forget gate, i t × tanh(W c [h t-1, z t + b c ) is the update of the memory cell by the current input z t under the control of the input gate, and tanh is the hyperbolic tangent activation function, ensuring that the new memory content is restricted within an appropriate range;

[0098] Execute the output gate, o t = σ × (Wo × [h t-1 , z t + b o ), where o t is the activation value of the output gate, which determines which information in the memory cell c t will affect the final output. W o is the weight matrix of the output gate, and b o is the bias term of the output gate;

[0099] Hidden state, the output h of the LSTM t is controlled by the output gate to affect the memory cell c t and is calculated through the tanh activation function. The formula is h t = o t × tanh(c t );

[0100] Obtain the LSTM output function y = soft max(W f × h t + b f ), where W f is the weight matrix of the forget gate, and b f is the bias term of the forget gate.

[0101] By further refining the working mechanism of the LSTM model, the system's ability to process time-series signals is improved. The design of each gate in the LSTM can flexibly adjust the information flow in the memory cell, enabling the system to dynamically update the memory content according to the characteristics of the input signal. The combined use of the forget gate and the input gate ensures that the system can efficiently update and retain important information, avoiding interference from redundant or irrelevant information. During the update process of the memory cell, the use of the tanh activation function ensures that the update amplitude of the memory content is within a reasonable range, avoiding excessive or too small changes, enabling the system to maintain stable performance. Through the action of the output gate, the LSTM can output an accurate prediction of the signal according to the current state. The design of the LSTM enables the system to flexibly adjust its internal state when facing complex time-series signals, accurately capture long-term dependencies, and thus provide stable and accurate signal processing capabilities in complex environments. By optimizing the update process of the memory cell, the system not only improves its ability to capture long-term signal dependencies but also enhances its signal processing efficiency and prediction accuracy, further enhancing the system's adaptability.

[0102] Execute an adaptive sensitivity adjustment strategy to automatically adjust the sensitivity of the detector based on environmental noise and detection distance, including:

[0103] Calculate the sensitivity Among them, S(t) represents the sensitivity at time t, S0 represents the initial sensitivity, and α represents the adjustment coefficient;

[0104] Correspondingly, when the environmental noise σ(t) increases, the sensitivity of the detector increases;

[0105] When the detection distance d(t) increases, the sensitivity of the detector increases.

[0106] Through the adaptive sensitivity adjustment strategy, the detector can automatically adjust the sensitivity according to the changes in environmental noise and detection distance, thereby optimizing the detection performance. When the environmental noise is large, the detector can capture weak signals by increasing the sensitivity, thus improving the detection accuracy; while when the detection distance is far, the detector expands the sensitivity range to ensure that the system can maintain sufficient detection ability. The automatic adjustment of sensitivity not only improves the adaptability of the system in different environments but also effectively copes with the impact of external interference. By calculating and adjusting the sensitivity in real time, the system can maintain a high detection accuracy in complex environments, avoiding the limitations of fixed sensitivity settings. The introduction of the sensitivity adjustment mechanism improves the reliability of the system in dynamic environments, ensuring that the system can automatically optimize the detection performance according to the current environmental conditions, thereby ensuring the accuracy and stability of the signal.

[0107] Perform Fourier frequency-domain analysis strategy on the time data sequence signal, predict the time data sequence signal at future moments, and based on the predicted time data sequence signal, judge whether the environmental noise affects the signal, including:

[0108] Obtain the time data sequence signal x(t);

[0109] Use Fourier transform to convert the signal from the time domain to the frequency domain Among them, X(f) is the frequency-domain signal, x(t) is the time-domain signal, and f is the frequency;

[0110] Set the frequency interval;

[0111] Judge whether the frequency-domain signal obtained by Fourier transform is within the frequency interval. If it exists, it is determined that the environmental noise does not affect the signal;

[0112] If it does not exist, it is determined that the environmental noise affects the signal;

[0113] When the environmental noise affects the signal, execute the environmental filtering adaptive noise reduction strategy to reduce the interference of the environmental noise. Specifically:

[0114] Add the time data sequence signal and the environmental noise to the filter, then the size of the signal received by the filter is y(t) = x(t) + σ(t); the size of the signal output by the filter is Among them, * represents the convolution operation, and w(t) is the weight vector of the adaptive filter;

[0115] Calculate the filtering and noise reduction error Update the filter weights using the minimum filtering mean square error, specifically:

[0116] Calculate the squared mean value of the filtering and noise reduction error:

[0117] Among them, E(·) represents the expected value of the filtering and noise reduction error;

[0118] Update the weight vector w(t + 1) = w(t) + μ × e(t) × y(t), where μ is the learning rate that controls the step size of weight update, and e(t) is the current error;

[0119] Use the updated weight vector to calculate the new filtering and noise reduction error and update the squared mean value of the filtering and noise reduction error;

[0120] Set an error threshold, compare the squared mean value of the filtering and noise reduction error with the error threshold. If the filtering and noise reduction error is less than or equal to the error threshold, stop updating the filter weights.

[0121] Through the Fourier frequency domain analysis strategy, the system can effectively analyze the spectral characteristics of the signal, thereby judging whether the environmental noise affects the signal. The Fourier transform can convert the time-domain signal into a frequency-domain signal, making the frequency components of the signal more intuitive. Especially when the signal contains multiple frequency components, frequency domain analysis can help the system accurately identify the influence range of the noise. By setting the frequency interval, the system can judge whether the signal is affected by noise and perform corresponding processing according to the analysis results. If the noise affects the signal, the system will perform noise reduction through the adaptive filtering strategy to reduce the interference of the noise on the signal. The dynamic weight update of the filter and the minimum mean square error optimization algorithm make the noise reduction process more efficient and accurate. This strategy can maintain high-efficient signal processing ability in different noise environments. Especially in a noisy environment, it can effectively improve the quality and accuracy of the signal. Through Fourier frequency domain analysis and adaptive noise reduction, the system can significantly improve the reliability of the signal and avoid the influence of noise on the signal analysis and prediction results.

[0122] Example 2, refer to Figure 2 , A handheld electronic component inspection and analysis system, which is applied to the above-mentioned handheld electronic component inspection and analysis method, includes:

[0123] Signal acquisition and multi-band analysis module: Perform multi-band analysis through the fundamental wave signal in the S band and its harmonic reflection signals;

[0124] Convolution feature extraction module: Extract the fundamental wave and harmonic features in the signal through convolution and pooling;

[0125] Time-sequence dependence capture module: The LSTM network captures the time-sequence dependence in the signal for long-term signal prediction;

[0126] Sensitivity adjustment module: Adaptively adjusts the detector sensitivity to adapt to different environmental noises and detection distances;

[0127] Fourier analysis and noise reduction module: Conducts frequency-domain analysis through Fourier transform and combines with adaptive filtering to reduce environmental noise interference.

[0128] Embodiment 3, a handheld electronic component inspection and analysis device applied to the above-mentioned handheld electronic component inspection and analysis method, includes:

[0129] Use a high-speed data bus to transmit signals. The radio frequency transmitter sends the fundamental wave signal. The receiving module receives the signal reflected by the target object, transmits the received signal to the digital signal processing unit after analog-to-digital conversion. The digital signal processing unit processes the signal and then passes it to the deep learning acceleration unit for feature extraction and time-sequence analysis. The signal output by the deep learning unit is used in cooperation with the Fourier analysis and noise reduction module to judge the influence of environmental noise. The environmental monitoring and sensing device collects environmental data and then passes it to the sensitivity adjustment module to adjust the detector sensitivity, and interacts with the user through the user interface.

[0130] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0131] The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A handheld electronic component inspection and analysis method, characterized in that: include: The transmitter sends the fundamental wave of the S band to the target object; The fundamental wave reflected by the target object and the second harmonic and third harmonic generated by the target object are captured by the receiving end; Get the time data series signal x(t) received by the receiving end = {x1, x2, ..., x t }, where x i Represents the discretized time data series signal, t is the time step number, and t is a positive integer; Execute the multi-convolution feature extraction strategy to extract the fundamental wave feature z1, the second harmonic feature z2 and the third harmonic feature z3 in the time data series signal, and obtain the feature map Z = {z1, z2, z3} of the time step sequence; According to the feature graph of the time step sequence, the timing dependency capture strategy is executed to capture the dependency relationship of the time data series signal; Obtain the ambient noise σ(t) and detection distance d(t) at time t; Execute adaptive sensitivity adjustment strategy to automatically adjust the sensitivity of the detector based on environmental noise and detection distance; Calculated sensitivity Where S(t) represents the sensitivity at time t, S0 represents the initial sensitivity, and α represents the adjustment coefficient; Correspondingly, when the environmental noise σ(t) increases, the sensitivity of the detector increases; When the detection distance d(t) increases, the sensitivity of the detector increases; Performing Fourier frequency domain analysis on the time data series signal to predict the time data series signal at a future moment, and judging whether the environmental noise affects the signal based on the predicted time data series signal; If the environmental noise affects the signal, the environmental filtering adaptive noise reduction strategy is executed to reduce the interference of the environmental noise; Get the time data series signal x(t); Convert the signal from the time domain to the frequency domain using Fourier transform Where X(f) is the frequency domain signal, x(t) is the time domain signal, and f is the frequency; Set the frequency range; Determine whether the frequency domain signal obtained by Fourier transform is within the frequency interval. If so, it is determined that the environmental noise does not affect the signal; If it does not exist, it is considered that the environmental noise affects the signal; When environmental noise affects the signal, the environmental filtering adaptive noise reduction strategy is executed to reduce the interference of environmental noise. Specifically: When the time data series signal and environmental noise are added to the filter, the signal size received by the filter is y(t)=x(t)+σ(t); The signal size of the filter output is Where * is the convolution operation, w(t) is the weight vector of adaptive filtering; Calculate the filtering noise reduction error Use the minimum filtered mean square error to update the filter weights, specifically: Calculate the squared mean of the filtering noise reduction error: Where E(·) represents the expected value of the filtering noise reduction error; Update the weight vector w(t+1)=w(t)+μ×e(t)×y(t), where μ is the learning rate, which controls the step size of the weight update, and e(t) is the current error; The updated weight vector is used to calculate the new filtering noise reduction error and update the square average of the filtering noise reduction error; Set an error threshold and compare the square average of the filtering noise reduction error with the error threshold. If the filtering noise reduction error is less than or equal to the error threshold, stop updating the filter weights.

2. The handheld electronic component inspection and analysis method according to claim 1 is characterized in that: The multi-convolution feature extraction strategy is executed to extract the fundamental wave feature z1, the second harmonic feature z2 and the third harmonic feature z3 in the time data series signal, including: Perform a convolution operation: Set the fundamental frequency to f1; Then the frequency of the second harmonic is f2=2f1, and the frequency of the third harmonic is f3=3f1; Perform convolution operation on the fundamental wave to obtain the fundamental wave feature map Perform convolution operation on the second harmonic to obtain the second harmonic feature map Perform convolution operation on the third harmonic to obtain the third harmonic feature map Among them, x i+j-1 is the discrete value of the time data series signal at the i+j-1th moment, w 1j ,w 2j ,w 3j are the jth weights of the convolution kernels for the fundamental, second harmonic, and third harmonic, respectively; b1, b2, and b3 are the bias terms for the fundamental, second harmonic, and third harmonic, respectively; Perform pooling operations: Max pooling operation Where i = 1, 2, 3, z i is the output of the convolutional layer, pool_region is the area of the pooling operation window, p i is the output result after pooling.

3. The handheld electronic component inspection and analysis method according to claim 1, characterized in that: The method of executing a timing dependency capture strategy based on the feature graph of the time step sequence to capture the dependency relationship of the time data sequence signal includes: Execute the input gate, which controls the current input z t and the hidden state of the previous time step and h t-1 How to update the contents of a memory cell t =σ×(W i ×[h t-1 ,z t ]+b i ), where i t is the activation value of the input gate, which determines which parts of the current input will be written into the memory unit; σ is the sigmoid activation function, and the output range is between [0, 1], W i is the weight matrix of the input gate, b i is the bias term of the input gate; Execute the forget gate, f t =σ×(W f ×[h t-1 ,z t ]+b f ), where f t is the activation value of the forget gate, which controls how much the content stored in the memory unit is forgotten. f is the weight matrix of the forget gate, b f is the bias term of the forget gate.

4. The handheld electronic component inspection and analysis method according to claim 3 is characterized in that: The method of executing a timing dependency capture strategy based on the feature graph of the time step sequence to capture the dependency relationship of the time data sequence signal includes: Update memory unit, c t =f t ×c t-1 +i t ×tanh(W c [h t-1, z t ]+b c ), where f t ×c t-1 Represents the information retained from the previous moment’s memory determined by the forget gate, i t ×tanh(W c [h t-1, z t ]+b c ) is the input gate control, the current input z t For the update of memory units, tanh is the hyperbolic tangent activation function, which ensures that the new memory content is limited to an appropriate range; Execute the output gate, o t =σ×(W o ×[h t-1 ,z t ]+b o ), where o t is the activation value of the output gate, which determines the memory unit c t Which information will affect the final output, W o is the weight matrix of the output gate, b o is the bias term of the output gate; Hidden state, LSTM output h t The memory unit c is controlled by the output gate t The influence of , and it is calculated by the tanh activation function, the formula is h t =o t ×tanh(c t ); Get the LSTM output function y = softmax(W f ×h t +b f ), where W f is the weight matrix of the forget gate, b f is the bias term of the forget gate.

5. A handheld electronic component inspection and analysis system, applied to the handheld electronic component inspection and analysis method according to any one of claims 1 to 4, characterized in that: include: Signal acquisition and multi-band analysis module: performs multi-band analysis through the S-band fundamental signal and its harmonic reflection signal; Convolutional feature extraction module: extracts fundamental and harmonic features from the signal through convolution and pooling; Temporal Dependency Capture Module: The LSTM network captures temporal dependencies in signals and performs long-term signal prediction, specifically including: Calculated sensitivity Where S(t) represents the sensitivity at time t, S0 represents the initial sensitivity, and α represents the adjustment coefficient; Correspondingly, when the environmental noise σ(t) increases, the sensitivity of the detector increases; When the detection distance d(t) increases, the sensitivity of the detector increases; Sensitivity adjustment module: adaptively adjusts the detector sensitivity to adapt to different environmental noise and detection distances; Fourier analysis and noise reduction module: This module uses Fourier transform to perform frequency domain analysis and combines adaptive filtering to reduce environmental noise interference. Specifically, it includes: Get the time data series signal x(t); Convert the signal from the time domain to the frequency domain using Fourier transform Where X(f) is the frequency domain signal, x(t) is the time domain signal, and f is the frequency; Set the frequency range; Determine whether the frequency domain signal obtained by Fourier transform is within the frequency interval. If so, it is determined that the environmental noise does not affect the signal; If it does not exist, it is considered that the environmental noise affects the signal; When environmental noise affects the signal, the environmental filtering adaptive noise reduction strategy is executed to reduce the interference of environmental noise. Specifically: When the time data series signal and environmental noise are added to the filter, the signal size received by the filter is y(t)=x(t)+σ(t); The signal size of the filter output is Where * is the convolution operation, w(t) is the weight vector of adaptive filtering; Calculate the filtering noise reduction error Use the minimum filtered mean square error to update the filter weights, specifically: Calculate the squared mean of the filtering noise reduction error: Where E(·) represents the expected value of the filtering noise reduction error; Update the weight vector w(t+1)=w(t)+μ×e(t)×y(t), where μ is the learning rate, which controls the step size of the weight update, and e(t) is the current error; The updated weight vector is used to calculate the new filtering noise reduction error and update the square average of the filtering noise reduction error; Set an error threshold and compare the square average of the filtering noise reduction error with the error threshold. If the filtering noise reduction error is less than or equal to the error threshold, stop updating the filter weights.

6. A handheld electronic component inspection and analysis device, applied to the handheld electronic component inspection and analysis method according to any one of claims 1 to 4, characterized in that: include: A high-speed data bus is used to transmit signals, and the radio frequency transmitter sends the fundamental wave signal. The receiving module receives the signal reflected by the target object and transmits the received signal to the digital signal processing unit after analog-to-digital conversion. The digital signal processing unit processes the signal and then passes it to the deep learning acceleration unit for feature extraction and timing analysis. The signal output by the deep learning unit is used in conjunction with the Fourier analysis and noise reduction module to determine the impact of environmental noise. The environmental monitoring and sensing equipment collects environmental data and passes it to the sensitivity adjustment module to adjust the detector sensitivity, and interacts with the user through the user interface.

Images