High-frequency electrocardio waveform analysis method and system based on image recognition

Abstract

The invention relates to the technical field of electrocardiosignal processing and image recognition, and discloses a high-frequency electrocardiosignal waveform analysis method and system based on image recognition, which realize simulation domain pre-correction processing, adaptive parameter adjustment and subsequent image recognition analysis of original simulation electrocardiosignals by setting a detection end, an adjustment end and a recognition end. Therefore, a complete and efficient analysis system is formed, the problem that the signal quality is influenced by equipment difference in the prior art is effectively solved, and the accuracy and reliability of overall analysis are improved.

Description

Technical Field

[0001] This invention relates to the fields of electrocardiogram signal processing and image recognition technology, and in particular to a high-frequency electrocardiogram waveform analysis method and system based on image recognition. Background Technology

[0002] In the auxiliary diagnosis of cardiovascular diseases, high-frequency electrocardiogram (ECG) waveform analysis is an important technique. It analyzes waveform features by converting ECG signals into images and performing image recognition to detect subtle lesions that are difficult to detect with traditional ECGs. However, in practical applications, especially in primary healthcare institutions, the variety and age of equipment means that the raw ECG signals are easily interfered with during acquisition and conversion, leading to a decline in signal quality. This severely affects the accuracy of subsequent image recognition, making early warnings unreliable.

[0003] Specifically, existing high-frequency electrocardiogram (ECG) waveform analysis systems based on image recognition rely on a highly standardized and homogeneous ECG image dataset for training and optimization. This dataset originates from medical institutions in a specific region, and all ECG signals are acquired using standardized, rigorously calibrated ECG acquisition equipment. This training method ensures that, under ideal conditions, the system can accurately identify and analyze minute, continuously changing morphological features in high-frequency ECG waveforms that indicate myocardial ischemia or arrhythmia.

[0004] However, in primary healthcare institutions, due to cost control and equipment utilization pressures, the ECG acquisition equipment used daily exhibits significant diversity, including older models and relatively basic portable devices. The hardware specifications, internal circuit design, and signal processing capabilities of these devices differ significantly from the standardized equipment initially used for system training. For example, older equipment may introduce power frequency interference due to degraded shielding performance, causing baseline drift in the original signal; portable devices may use lower sampling rates, leading to attenuation or distortion of key high-frequency components in the high-frequency QRS signal. These signal variations caused by equipment differences have a range and patterns far exceeding the scope covered by the initial training dataset.

[0005] Faced with raw ECG signals from complex sources and of varying quality, the limitations of the standard "signal-to-image conversion function" and "image normalization processing module" in existing systems become apparent. The internal parameters and processing logic of these modules are rigidly designed for the characteristics of the initial standardized signal, lacking adaptive adjustment capabilities. When processing non-standard signals, their fixed processing logic cannot effectively distinguish between genuine pathological information and noise introduced by the equipment. As a result, some artifacts and interference introduced by equipment differences may be incorrectly preserved or even amplified during image conversion, creating misleading features in the generated image. Simultaneously, some subtle waveform details that truly characterize early myocardial ischemia or arrhythmia may be incorrectly treated as noise and smoothed because their morphology does not conform to the "standard," causing this crucial information to be erased from the image. Ultimately, the quality of the generated ECG waveform image used for image recognition and analysis is severely degraded, and its information fidelity is significantly reduced.

[0006] This information loss or distortion during image conversion directly weakens the convolutional neural network's ability to accurately learn and recognize image features. While the system's core weight matrix performs well on standardized data, its sensitivity and robustness to features significantly decrease when faced with contaminated or distorted images. In particular, subtle, continuously changing morphological features in high-frequency QRS waveforms that indicate myocardial ischemia or arrhythmia become blurred in these low-quality images or are mixed with interference introduced by the device, making it difficult for the system to accurately extract and distinguish them. This directly leads to a significantly higher misclassification rate in practical applications, especially when identifying early lesions with only subtle changes, severely impacting the reliability of its diagnosis.

[0007] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0008] This invention provides a high-frequency electrocardiogram waveform analysis method and system based on image recognition, aiming to solve the problem in the prior art where the quality of the original electrocardiogram signal deteriorates due to the diversity of equipment in primary medical institutions, which in turn affects the accuracy of image recognition and makes early warning unreliable.

[0009] In a first aspect, to address the aforementioned technical problems, the present invention provides a high-frequency electrocardiogram waveform analysis method based on image recognition, comprising: Receive raw analog electrocardiogram signals; The original analog ECG signal is subjected to analog domain pre-correction processing, which includes: real-time monitoring of the signal characteristics of the original analog ECG signal, including baseline position and intensity of rapidly changing components; Based on the signal characteristics, the simulation parameters of the original simulated ECG signal are adaptively adjusted, including the baseline position and the gain of rapidly changing components. The simulated electrocardiogram signal, after being pre-corrected in the analog domain, is used for image recognition and analysis.

[0010] This technical solution enables real-time monitoring of signal characteristics and adaptive adjustment of analog parameters by pre-correcting the original electrocardiogram signal in the analog domain. This effectively solves the problem of signal quality degradation caused by equipment differences and significantly improves the accuracy of subsequent image recognition and the reliability of early warning.

[0011] Furthermore, based on the above, the analog parameters of the original simulated ECG signal are adaptively adjusted according to the signal characteristics. These analog parameters include baseline position and gain of rapidly changing components, including: The original analog ECG signal is split into the first analog processing path and the second analog processing path; According to the first simulation processing path, a variable cutoff frequency high-pass filter is used to perform high-frequency background purification on the original analog ECG signal. According to the second simulation processing path, a bandpass filter bank is used to extract low-frequency physiological features from the original simulated electrocardiogram signal. The output signals of the first and second analog processing paths are simulated and fused together to reconstruct the original analog ECG signal, thereby adaptively adjusting the analog parameters of the original analog ECG signal, including baseline position and gain of rapidly changing components.

[0012] This technical solution allows for more precise adaptive adjustment of simulation parameters by splitting the signal into two independent analog processing paths, performing high-frequency background purification and low-frequency physiological feature extraction, and then performing analog fusion reconstruction. This enables more effective recovery and optimization of ECG signal quality under complex interference environments, providing a purer and more accurate signal source for subsequent analysis.

[0013] In some preferred embodiments, according to the second analog processing path, a bandpass filter bank is used to extract low-frequency physiological features from the original analog electrocardiogram signal, including: Spectral analysis was performed on the original simulated electrocardiogram signal to determine the baseline drift interference characteristics in the low-frequency band; Based on the baseline drift interference characteristics, the cutoff frequency or quality factor of at least one filter in the bandpass filter bank is dynamically adjusted to suppress the baseline drift interference. For the signal processed by the bandpass filter bank, the signal components within the preset low-frequency band are extracted as the low-frequency physiological features; The passband gain of the bandpass filter bank is adaptively fine-tuned based on the amplitude changes of the low-frequency physiological characteristics.

[0014] This technical solution enables the more accurate extraction of low-frequency physiological characteristics and effective suppression of low-frequency interference by continuously monitoring low-frequency interference characteristics and dynamically adjusting the filtering parameters of the bandpass filter bank, as well as fine-tuning the amplitude and phase of low-frequency physiological characteristics, thereby ensuring the integrity and accuracy of key physiological information.

[0015] Furthermore, the output signals of the first analog processing path and the second analog processing path are subjected to analog fusion reconstruction, including: Continuously monitor the instantaneous phase difference and instantaneous amplitude ratio between the output signals of the first analog processing path and the output signals of the second analog processing path; Adjust the simulation delay based on the instantaneous phase difference; Adjust the programmable gain according to the instantaneous amplitude ratio; The output signal of the adjusted first analog processing path is superimposed with the output signal of the adjusted second analog processing path.

[0016] Through this technical solution, this application continuously monitors the instantaneous phase difference and instantaneous amplitude ratio, and adjusts the analog delay and programmable gain accordingly before performing analog superposition. This enables precise synchronization and optimized fusion of two path signals, thereby preserving effective information and suppressing noise to the greatest extent during the reconstruction process, and improving the quality of the fused signal.

[0017] Based on the above, the low-frequency interference characteristics of the original simulated electrocardiogram signal were continuously monitored, including: Continuously monitor the instantaneous energy and instantaneous frequency change rate of the low-frequency component of the original analog electrocardiogram signal; Determine whether the instantaneous energy and instantaneous frequency change rate exceed preset thresholds; The starting and ending boundaries of the disturbance are identified based on the rising and falling edges of the instantaneous energy. The frequency type of interference is determined based on the instantaneous energy.

[0018] Through this technical solution, this application continuously monitors the instantaneous energy and instantaneous frequency change rate of the low-frequency component, determines whether it exceeds a preset threshold, and identifies the interference boundary and determines the frequency type based on the instantaneous energy. This enables refined identification and classification of low-frequency interference, providing an accurate basis for subsequent dynamic filtering parameter adjustments, thereby more effectively suppressing interference.

[0019] As a technological improvement, the transient energy of the low-frequency component of the original analog electrocardiogram signal is continuously monitored, including: The low-frequency component of the original analog ECG signal is input into multiple parallel analog square-law detectors, so that each analog square-law detector performs energy integration on the low-frequency component of the original analog ECG signal according to different time constants, resulting in multiple integration results. The instantaneous energy of the low-frequency component is obtained by dynamically weighting and averaging multiple integral results.

[0020] This technical solution utilizes multiple parallel analog square-law detectors to integrate energy and dynamically weights the integration results, enabling a more comprehensive and accurate capture of instantaneous energy changes in the low-frequency range, thereby improving the sensitivity and robustness of energy monitoring.

[0021] As a further improvement, the instantaneous frequency change rate of the low-frequency component of the original analog electrocardiogram signal is continuously monitored, including: The low-frequency portion of the original analog electrocardiogram signal is input into multiple parallel analog frequency tracking units, so that each analog frequency tracking unit is preset with a specific frequency capture range and tracking bandwidth. It tracks and captures instantaneous frequencies within the range in real time and outputs the corresponding instantaneous frequency signals; When the instantaneous frequency signals of multiple analog frequency tracking units cross or merge, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted according to the signal strength and frequency change trend of each analog frequency tracking unit to ensure that the instantaneous frequency of each independent interference source is continuously tracked. The instantaneous frequency signal output by each analog frequency tracking unit is processed by analog differentiation to obtain the instantaneous frequency change rate of each independent interference source.

[0022] Through this technical solution, this application uses multiple parallel analog frequency tracking units to track instantaneous frequencies in real time and dynamically adjusts tracking parameters when frequencies cross or merge. This ensures continuous and accurate monitoring of the instantaneous frequency change rate of complex multi-source low-frequency interference, providing refined frequency information for interference suppression.

[0023] To improve the solution, when the instantaneous frequency signals of multiple analog frequency tracking units intersect or merge, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted based on the signal strength and frequency change trends of each unit. This ensures that the instantaneous frequency of each independent interference source is continuously tracked, including: The signal strength of each analog frequency tracking unit is normalized. Based on the normalized signal strength and frequency variation trend, a weighting factor is dynamically assigned to each analog frequency tracking unit. Based on the weighting factor, increase the weight of the analog frequency tracking unit with low signal strength; Based on the normalized signal strength and frequency variation trend, the instantaneous frequency of each analog frequency tracking unit is compared in real time through an analog comparator array. Identify frequency crossover or fusion points; At frequency crossover or merging points, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted based on the weighted signal strength and frequency change trend of the analog frequency tracking unit.

[0024] This technical solution effectively solves the tracking problem when multiple interference frequencies cross or merge by normalizing the signal strength and dynamically allocating weighting factors, and by adjusting the tracking parameters at frequency crossover or fusion points according to the weighted signal strength and frequency change trends, ensuring that the instantaneous frequency of each independent interference source is accurately and continuously tracked.

[0025] To enhance functionality, the signal strength of each analog frequency tracking unit is normalized, including: Logarithmically compress the signal strength of each analog frequency tracking unit to obtain a logarithmically compressed signal; then differentially process the logarithmically compressed signal to obtain a differential signal. The gain of the programmable gain amplifier is dynamically adjusted based on the instantaneous amplitude of the differential signal. The differential signal after gain adjustment is normalized.

[0026] This technical solution effectively broadens the dynamic range of signal strength and improves the accuracy and robustness of normalization by performing logarithmic compression, differential processing, dynamic adjustment of programmable gain based on instantaneous amplitude, and finally normalization processing on the signal strength. This provides more reliable signal strength information for subsequent weight allocation and frequency tracking.

[0027] Secondly, this application also discloses a high-frequency electrocardiogram waveform analysis system based on image recognition, comprising: The detection end is used to receive the raw analog ECG signal; the raw analog ECG signal is subjected to analog domain pre-correction processing, which includes: real-time monitoring of the signal characteristics of the raw analog ECG signal, including baseline position and intensity of rapidly changing components; The adjustment terminal is used to adaptively adjust the analog parameters of the original analog ECG signal according to the signal characteristics. The analog parameters include the baseline position and the gain of the rapidly changing components. The recognition end is used to apply the simulated electrocardiogram signal, which has been pre-corrected in the analog domain, to image recognition and analysis.

[0028] Through this technical solution, this application achieves analog domain pre-correction processing, adaptive parameter adjustment, and subsequent image recognition analysis of the original analog electrocardiogram signal by setting up a detection end, an adjustment end, and an identification end, forming a complete and efficient analysis system. This effectively solves the problem of signal quality being affected by equipment differences in the prior art, and improves the accuracy and reliability of the overall analysis.

[0029] Beneficial effects This application provides a high-frequency electrocardiogram (ECG) waveform analysis method based on image recognition. By pre-correcting the raw analog ECG signal in the analog domain, it can monitor signal characteristics in real time, including baseline position and the intensity of rapidly changing components, and adaptively adjust analog parameters, such as baseline position and the gain of rapidly changing components, according to these characteristics. This adaptive pre-correction processing in the analog domain effectively solves the problem in existing technologies where the raw ECG signal is easily interfered with during acquisition and conversion due to the diversity and varying age of equipment in primary healthcare institutions, resulting in decreased signal quality and affecting the accuracy of subsequent image recognition. By optimizing the signal directly in the analog domain, it avoids the quantization errors and information loss that may be introduced by digital domain processing, ensuring the fidelity of key minute features in the high-frequency ECG waveform. The pre-corrected analog ECG signal is used for image recognition analysis, significantly improving the accuracy and reliability of image recognition, greatly enhancing the reliability of early warning and diagnosis, and overcoming the shortcomings of existing technologies where signal contamination or distortion leads to increased misjudgment rates and compromised diagnostic reliability. Attached Figure Description

[0030] Figure 1 This is a flowchart of a high-frequency electrocardiogram waveform analysis method based on image recognition provided in an embodiment of the present invention; Figure 2 This is a flowchart of a method for adaptively adjusting analog parameters based on signal characteristics, provided by an embodiment of the present invention. Figure 3 This is a flowchart of a method for extracting and processing raw analog electrocardiogram signals according to an embodiment of the present invention; Figure 4 This is a structural diagram of a high-frequency electrocardiogram waveform analysis system based on image recognition provided in an embodiment of the present invention. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Reference Figure 1 , Figure 1 This is a flowchart of a high-frequency electrocardiogram waveform analysis method based on image recognition provided in an embodiment of the present invention, including: S1 receives the raw analog electrocardiogram signal; S2, Perform analog domain pre-correction processing on the original simulated ECG signal, the analog domain pre-correction processing includes: real-time monitoring of the signal characteristics of the original simulated ECG signal, the signal characteristics including baseline position and intensity of rapidly changing components; S3, based on the signal characteristics, adaptively adjust the simulation parameters of the original simulated electrocardiogram signal, the simulation parameters including baseline position and gain of rapidly changing components; S4, the simulated electrocardiogram signal after analog domain pre-correction processing is used for image recognition analysis.

[0033] This application introduces analog domain pre-correction processing, which can monitor the signal characteristics of the original analog ECG signal in real time and adaptively adjust the analog parameters according to these characteristics, thereby effectively improving the signal quality and providing a more reliable data foundation for subsequent image recognition and analysis. This significantly improves the limitations of traditional methods in processing non-standardized ECG signals.

[0034] To better understand the technical solution proposed in this application, some key terms involved will be explained first.

[0035] "Raw analog electrocardiogram signal" refers to the raw electrophysiological signal collected from the human body without any digital processing. It contains rich information about heart activity, but is also susceptible to various noises and interferences.

[0036] "Analog domain pre-calibration processing" refers to the preprocessing of analog signals before digitization, which aims to improve signal quality, eliminate or reduce interference, and make them more suitable for subsequent analysis.

[0037] "Signal characteristics" refer to the inherent properties exhibited by the raw analog electrocardiogram signal, such as its baseline position and the intensity of rapidly changing components. These characteristics are important for assessing signal quality and guiding pre-correction processing.

[0038] "Baseline position" refers to the average potential of the electrocardiogram waveform in the vertical direction. Ideally, it should remain stable, but it often shifts due to breathing, body movement, or equipment drift.

[0039] "Intensity of rapidly changing components" refers to the activity level of high-frequency components in an electrocardiogram (ECG) signal. These high-frequency components are usually associated with rapid electrophysiological events such as the QRS complex, and their intensity changes can reflect the transient characteristics of the signal.

[0040] "Analog parameters" refer to the adjustable parameters in analog domain pre-calibration processing, such as baseline position and gain of rapidly changing components. By adjusting these parameters, fine control of the signal can be achieved.

[0041] The “baseline position” is a simulation parameter whose adjustment aims to restore the baseline of the electrocardiogram waveform to a preset reference potential and eliminate baseline drift.

[0042] "Rapidly changing component gain" is an analog parameter whose adjustment is designed to control the amplification or attenuation of rapidly changing components in the electrocardiogram signal in order to optimize its performance in image recognition.

[0043] "Image recognition analysis" refers to converting pre-corrected analog electrocardiogram (ECG) signals into image form and using image processing and machine learning techniques to extract features and recognize patterns from ECG waveform images to assist in diagnosis.

[0044] The high-frequency electrocardiogram waveform analysis method based on image recognition proposed in this application is based on performing analog domain pre-correction processing on the original analog electrocardiogram signal to improve signal quality, thereby providing more reliable data for subsequent image recognition analysis.

[0045] Specifically, this method first receives raw analog electrocardiogram (ECG) signals. These analog signals can be acquired from electrodes on the patient's skin, for example, using ECG acquisition devices such as multi-lead electrocardiographs or portable ECG recorders. These signals are typically continuous voltage waveforms that directly reflect the heart's electrical activity.

[0046] Upon receiving the raw analog ECG signal, it undergoes analog domain pre-calibration processing. This processing includes real-time monitoring of the signal characteristics of the raw analog ECG signal. For example, the baseline position and intensity of rapidly changing components can be obtained in real time using peak and mean detectors in the analog circuit. The baseline position can be approximated by averaging the signal after low-pass filtering, while the intensity of rapidly changing components can be obtained by calculating their instantaneous energy or amplitude after high-pass filtering.

[0047] Based on the monitored signal characteristics, the analog parameters of the original simulated ECG signal are adaptively adjusted. These analog parameters include baseline position and gain of rapidly changing components. For example, when baseline drift is detected, an analog feedback loop compares the baseline position signal with a reference potential and uses the error signal to drive a programmable DC bias circuit, thereby adjusting the ECG signal baseline back to the preset reference potential. When the intensity of rapidly changing components is detected to be too low or too high, the gain can be adjusted using a programmable gain amplifier. For example, if the intensity of rapidly changing components is below a certain threshold, the gain is increased to amplify these components; if the intensity is too high, the gain is decreased to avoid saturation.

[0048] Finally, the analog ECG signals, after analog domain pre-calibration, are used for image recognition analysis. For example, the pre-calibrated analog signal can be converted into a digital signal via an analog-to-digital converter (ADC), and then converted into an ECG waveform image using digital signal processing techniques. These images can then be fed into a pre-trained convolutional neural network (CNN) or other image recognition model to identify specific morphological features in high-frequency QRS waveforms, thereby aiding in the diagnosis of cardiovascular diseases.

[0049] Overall, the high-frequency ECG waveform analysis method based on image recognition proposed in this application aims to address the challenges faced by traditional methods in processing non-standardized raw ECG signals. Traditional methods lack adaptive pre-correction capabilities for raw analog ECG signals, leading to significant interference and noise introduced by equipment differences in complex environments such as primary healthcare institutions, which severely impact signal quality and consequently reduce the accuracy of image recognition.

[0050] This application achieves real-time, adaptive optimization of raw analog ECG signals by introducing analog domain pre-correction processing. Specifically, firstly, the system receives raw analog ECG signals, which may suffer from baseline drift, high-frequency component attenuation, or distortion due to the diversity of acquisition devices. Next, these signals undergo pre-correction processing in the analog domain. The core of this processing lies in real-time monitoring of signal characteristics, including baseline position and the intensity of rapidly changing components. For example, through specially designed analog circuitry, baseline drift and intensity changes of rapid electrophysiological events such as QRS complexes can be continuously tracked.

[0051] Based on these real-time monitored signal characteristics, the system can adaptively adjust the analog parameters of the original analog ECG signal, such as baseline position and gain of rapidly changing components. When baseline drift is detected, the system dynamically adjusts the baseline position parameter to restore the signal baseline to a normal level, thereby eliminating low-frequency interference caused by respiration, body movement, or device drift. Simultaneously, when abnormal intensity of the rapidly changing component is detected, the system adjusts the gain of the rapidly changing component to ensure that key details in the high-frequency QRS waveform are appropriately amplified or attenuated, avoiding information loss or oversaturation. This adaptive adjustment mechanism allows the system to dynamically optimize signal quality according to the signal characteristics of different devices and real-time interference conditions, rather than using fixed processing logic.

[0052] Ultimately, the simulated ECG signal, after analog domain pre-calibration, exhibited a more stable baseline, clearer high-frequency components, and a significantly improved signal-to-noise ratio. These high-quality analog signals were then used for image recognition analysis. Due to the fundamental improvement in signal quality input to the image recognition module, artifacts and distortions introduced during image conversion were greatly reduced, and subtle waveform details that truly characterize pathological information were preserved and highlighted. This enabled subsequent image recognition models, such as convolutional neural networks, to more accurately learn and recognize morphological features in high-frequency QRS waveforms, significantly improving the reliability and accuracy of diagnosis, especially in identifying early and subtle lesions.

[0053] Compared with existing technologies, the core innovation of this application lies in its introduction of adaptive pre-correction processing in the analog domain. Traditional methods typically process signals after digitization, and their fixed processing logic is difficult to effectively cope with the complex interference and equipment differences that already exist in the analog signal during the acquisition stage. For example, when faced with power frequency interference introduced by old equipment or high-frequency component attenuation caused by portable equipment, the fixed "signal-to-image conversion function" and "image normalization processing module" of existing systems cannot adaptively adjust, resulting in the retention of interference or the erasure of key information.

[0054] This application improves signal quality at its source by monitoring signal characteristics (such as baseline position and intensity of rapidly changing components) in real time in the analog domain and adaptively adjusting analog parameters (such as baseline position and gain of rapidly changing components). This adaptive adjustment in the analog domain allows the system to intervene in the signal earlier and more directly, effectively suppressing baseline drift and amplifying or attenuating specific frequency components, thus adjusting the signal to its optimal state before digitization. This "correction before recognition" strategy significantly improves the information fidelity of ECG signals, enabling subsequent image recognition and analysis based on higher-quality images, thereby greatly improving the accuracy and reliability of diagnosis. Therefore, this application demonstrates significant progress and practicality in processing raw ECG signals from diverse devices.

[0055] In some embodiments described above, an adaptive adjustment of the analog parameters of the original analog ECG signal based on signal characteristics is proposed for analog domain pre-correction processing. However, in practical applications, the original analog ECG signal often contains multiple frequency components, such as low-frequency baseline drift and respiratory artifacts, and high-frequency electromyographic interference. If only a single adaptive adjustment mechanism is used, it may be difficult to effectively distinguish and process these interferences and physiological signals in different frequency ranges, which may lead to insufficient adjustment accuracy of the analog parameters, affecting the accurate extraction of high-frequency QRS waveforms and subsequent image recognition analysis.

[0056] In this regard, refer to Figure 2 , Figure 2This is a flowchart of a method for adaptively adjusting analog parameters based on signal characteristics provided in an embodiment of the present invention. S3 includes: S31, the original simulated ECG signal is split into the first simulated processing path and the second simulated processing path; S32, according to the first simulation processing path, a variable cutoff frequency high-pass filter is used to perform high-frequency background purification processing on the original simulated electrocardiogram signal; S33, according to the second simulation processing path, a bandpass filter bank is used to perform low-frequency physiological feature extraction processing on the original simulated electrocardiogram signal; S34, the output signal of the first simulation processing path and the output signal of the second simulation processing path are simulated and fused to reconstruct the signal, so as to adaptively adjust the simulation parameters of the original simulated ECG signal, the simulation parameters including the baseline position and the gain of the rapidly changing component.

[0057] Specifically, the raw analog ECG signal is split into a first analog processing path and a second analog processing path. This aims to decompose the raw signal into different frequency components for independent processing, thereby achieving more refined pre-correction. The first analog processing path mainly focuses on high-frequency components, while the second analog processing path focuses on low-frequency components.

[0058] In the first analog processing path, a variable cutoff frequency high-pass filter is used to perform high-frequency background purification on the original analog ECG signal. This high-pass filter can effectively filter out low-frequency noise and baseline drift while retaining high-frequency QRS waveform information. The variable cutoff frequency setting allows the filter to dynamically adjust its filtering range according to the signal characteristics monitored in real time, adapting to the characteristics of ECG signals from different individuals or under different physiological states. This achieves precise purification of the high-frequency background and avoids unnecessary attenuation or distortion of the QRS waveform.

[0059] In the second simulation processing path, a bandpass filter bank is used to extract low-frequency physiological features from the original simulated ECG signal. This bandpass filter bank is designed to accurately extract low-frequency physiological features from the ECG signal, such as P waves, T waves, and baseline drift. Bandpass filtering effectively isolates high-frequency noise and allows for focused analysis of the characteristics of low-frequency components. The use of the filter bank allows for more detailed analysis and extraction of low-frequency features across different frequency ranges; for example, different filtering parameters can be set for different types of low-frequency interference, such as respiratory artifacts and electromyographic interference.

[0060] Subsequently, the output signals from the first and second analog processing paths are fused and reconstructed. The purpose of this step is to intelligently integrate the independently processed high-frequency and low-frequency signal components to adaptively adjust the analog parameters of the original analog ECG signal. Through analog fusion reconstruction, the integrity of the high-frequency QRS waveform and the stability of the low-frequency baseline position can be comprehensively considered, thereby precisely adjusting the baseline position and the gain of rapidly changing components. This ensures that the pre-corrected ECG signal effectively removes interference while accurately reflecting the true shape of the ECG waveform.

[0061] This application's solution achieves refined processing of different frequency components by splitting the original analog ECG signal into two independent analog processing paths. Specifically, the first analog processing path focuses on high-frequency background purification using a variable cutoff frequency high-pass filter, effectively suppressing the influence of low-frequency interference on the high-frequency QRS waveform and ensuring the accuracy of rapidly changing components. Simultaneously, the second analog processing path focuses on low-frequency physiological feature extraction using a bandpass filter bank, accurately identifying and processing low-frequency interference such as baseline drift, providing a reliable basis for adaptive adjustment of the baseline position. This frequency-division processing strategy allows for independent yet coordinated adjustment of high-frequency and low-frequency analog parameters, avoiding the mutual interference or inconsistencies that may occur with a single adjustment mechanism. Finally, through analog fusion reconstruction, the optimized high-frequency and low-frequency information is intelligently integrated, enabling more accurate and robust adaptive adjustment of the baseline position and rapidly changing component gain of the original analog ECG signal, effectively solving the accuracy and distortion problems that may exist in traditional single adjustment mechanisms when processing complex ECG signals.

[0062] Through the above technical solution, this application can effectively separate and target different frequency components in the original analog ECG signal, significantly improving the accuracy and robustness of analog domain pre-correction. Specifically, high-frequency background purification through the first analog processing path can preserve the details of the high-frequency QRS waveform to the maximum extent while effectively suppressing low-frequency interference. Low-frequency physiological feature extraction through the second analog processing path can more accurately identify and correct baseline drift and other low-frequency artifacts. This divide-and-conquer and then fusion strategy makes the adaptive adjustment of baseline position and rapidly changing component gain more precise, avoiding signal distortion or loss of key information that may be caused by single processing in traditional methods. As a result, the pre-corrected analog ECG signal has a higher signal-to-noise ratio and a more realistic waveform, greatly improving the accuracy and reliability of subsequent image recognition and analysis, and providing a solid foundation for the accurate diagnosis of high-frequency ECG waveforms.

[0063] In some preferred embodiments, it is assumed that the received raw analog ECG signal contains significant baseline drift (e.g., due to patient breathing or body movement) and some degree of electromyographic interference (high-frequency noise). First, the raw analog ECG signal is shunted. A first analog processing path receives the signal and processes it using a variable cutoff frequency high-pass filter. This high-pass filter dynamically adjusts its cutoff frequency based on the real-time monitored baseline drift intensity; for example, when the baseline drift is large, the cutoff frequency is appropriately increased to more effectively remove low-frequency components, thereby performing high-frequency background purification on the high-frequency QRS waveform.

[0064] Simultaneously, the second analog processing path receives the same raw analog ECG signal and processes it using a bandpass filter bank. This bandpass filter bank can contain multiple parallel bandpass filters; for example, one filter focuses on extracting respiratory artifact features from 0.5Hz to 2Hz, and another focuses on extracting electromyographic interference features from 5Hz to 15Hz. By extracting these low-frequency physiological features, the characteristics of low-frequency interference can be accurately identified and quantified. Finally, the output signals of the first and second analog processing paths are fused and reconstructed. During the fusion process, the baseline position of the high-frequency signal can be precisely adjusted based on the amplitude and phase information of the low-frequency physiological feature signals, and its gain can be adaptively adjusted based on the intensity of rapidly changing components of the high-frequency signal. For example, if the low-frequency path detects a significant baseline drift, the fusion module generates an inverse compensation signal superimposed on the output of the high-frequency path to correct the baseline. If the high-frequency path detects that the QRS waveform amplitude is too low or too high, the fusion module adjusts the overall gain. In this way, even in complex interference environments, the output analog ECG signal can be guaranteed to have a stable baseline and accurate QRS waveform amplitude, providing high-quality input for subsequent image recognition and analysis.

[0065] In some embodiments of this application, a bandpass filter bank is used to extract low-frequency physiological features from the original analog ECG signal via a second analog processing path. However, in practical applications, the original analog ECG signal is often affected by various low-frequency interferences, such as baseline drift, motion artifacts, or power line interference. Furthermore, the physiological characteristics of different individuals may vary. If the filtering parameters of the bandpass filter bank are fixed, the accuracy and robustness of low-frequency physiological feature extraction may decrease, making it unable to effectively cope with complex and changing signal environments, thereby affecting the reliability of subsequent image recognition and analysis.

[0066] To address this, this application further proposes an optimization scheme to improve the accuracy and adaptability of low-frequency physiological feature extraction. This scheme continuously monitors the low-frequency interference characteristics of the original analog electrocardiogram signal and dynamically adjusts the filtering parameters of the bandpass filter bank accordingly. Simultaneously, it monitors the amplitude and phase of the extracted low-frequency physiological features and fine-tunes the filtering parameters based on their changes.

[0067] For details, please refer to Figure 3 , Figure 3 This is a flowchart of a method for extracting and processing raw analog electrocardiogram signals according to an embodiment of the present invention, including the following steps: S331, Perform spectral analysis on the original analog electrocardiogram signal to determine the baseline drift interference characteristics in the low-frequency band; S332, Based on the baseline drift interference characteristics, dynamically adjust the cutoff frequency or quality factor of at least one filter in the bandpass filter bank to suppress the baseline drift interference; S333, for the signal processed by the bandpass filter bank, extract the signal components in the preset low-frequency band as the low-frequency physiological features; S334, adaptively fine-tuning the passband gain of the bandpass filter bank according to the amplitude change of the low-frequency physiological characteristics.

[0068] Specifically, the baseline drift interference characteristics are determined by simultaneously inputting the original analog ECG signal into BPF1. BPF1 is configured as a narrowband filter with a high Q value (e.g., Q=2) and a center frequency of 0.5Hz.

[0069] Characteristic Analysis: The output signal of BPF1 is detected using an analog RMS detection circuit or a precision rectification and low-pass filter circuit to obtain a DC voltage V_bias. The magnitude of this voltage V_bias is proportional to the baseline drift energy near 0.5Hz in the original signal, and is used as a quantitative indicator of the "baseline drift interference characteristic".

[0070] The filter parameters are dynamically adjusted to suppress interference. Specifically, the V_bias obtained in step A is compared with two preset thresholds (V_low, V_high), and a control signal is generated through a simple analog comparator logic or a micro analog controller (such as a simple state machine of a CPLD).

[0071] Dynamic adjustment strategy: If V_bias < V_low, it is determined that the baseline drift is slight. Control the digital potentiometer D1 to keep the center frequency of BPF1 at 0.5 Hz and the gain at -0.1. If V_low ≤ V_bias ≤ V_high, it is determined that the baseline drift is moderate. Control D1 to finely adjust the center frequency of BPF1 to 0.3 Hz and at the same time adjust its gain to -0.3 to enhance the cancellation of low-frequency drift.

[0072] If V_bias > V_high, it is determined that the baseline drift is severe. Control D1 to adjust the center frequency of BPF1 to 0.7 Hz, adjust the gain to -0.5, and slightly reduce the Q value of BPF1 (by adjusting another fixed resistor) to prevent oscillation to cope with severe drift.

[0073] Extract the physiological features in the preset low-frequency band. Specific implementation: After the original signal is preliminarily cancelled by BPF1 adjusted in step B, it is then input to BPF2. The output signal of BPF2 is the analog signal of the extracted low-frequency physiological features with baseline interference suppressed. Preset frequency band: The default passband of BPF2 is 0.67 - 5 Hz, which is determined by its center frequency of 1.2 Hz and Q value.

[0074] Fine-tune the passband gain according to the feature amplitude. Specific implementation: For the low-frequency physiological feature signal output by BPF2, obtain its envelope or average amplitude V_feat through another amplitude detection circuit. Adaptive fine-tuning: Compare this V_feat with an ideal reference amplitude V_ref. If V_feat continuously falls below V_ref (possibly due to individual differences or poor electrode contact resulting in weak feature signals), linearly increase the passband gain of BPF2 by controlling the digital potentiometer D3 (for example, increase by 1 dB each time) until V_feat enters the target range. If V_feat exceeds V_ref too much, correspondingly reduce the gain of BPF2 to prevent subsequent ADC saturation. This process is implemented through an analog integrator or a low-speed digital controller, and the adjustment speed is slow (time constant is about several seconds) to ensure stability.

[0075] Through the above technical solution, this application can effectively address the complex and variable low-frequency interference present in the original simulated electrocardiogram (ECG) signal, significantly improving the accuracy and robustness of low-frequency physiological feature extraction. Compared with traditional fixed-parameter filtering methods, this solution continuously monitors and dynamically adjusts the characteristics of low-frequency interference, and monitors and fine-tunes the amplitude and phase of low-frequency physiological features, ensuring stable and accurate acquisition of low-frequency physiological features of ECG signals under different physiological states and interference environments. This not only improves the extraction quality of low-frequency physiological features but also provides a purer and more reliable data foundation for the image recognition stage of high-frequency ECG waveform analysis, thereby improving the accuracy and reliability of the overall analysis method and reducing the risk of misdiagnosis or missed diagnosis.

[0076] Furthermore, the above-mentioned step of performing analog fusion reconstruction of the output signals of the first analog processing path and the second analog processing path can be further refined into the following operations: Continuously monitor the instantaneous phase difference and instantaneous amplitude ratio between the output signals of the first analog processing path and the output signals of the second analog processing path; Adjust the simulation delay based on the instantaneous phase difference; Adjust the programmable gain according to the instantaneous amplitude ratio; The output signal of the adjusted first analog processing path is superimposed with the output signal of the adjusted second analog processing path.

[0077] The phrase "continuously monitoring the instantaneous phase difference and instantaneous amplitude ratio between the output signals of the first and second analog processing paths" refers to acquiring the phase difference and amplitude ratio between the signals from the high-frequency background purification processing path and the low-frequency physiological feature extraction processing path in real time through dedicated analog phase detection and amplitude comparison circuits. For example, the instantaneous phase difference can be achieved using an analog phase detector, while the instantaneous amplitude ratio can be achieved using an analog divider or a logarithmic amplifier array. The purpose is to provide accurate real-time data for subsequent signal alignment and gain balancing.

[0078] "Adjusting the analog delay based on the instantaneous phase difference" refers to using a variable analog delay line to compensate for the timing of one signal (e.g., the output signal of the first analog processing path) based on the real-time monitored instantaneous phase difference, ensuring that the two signals are aligned in time. Specifically, the analog delay can be implemented using a voltage-controlled delay line or a current-controlled delay line, the purpose of which is to eliminate phase mismatch caused by differences in the propagation of different processing paths or physiological signals, ensuring the effectiveness of signal superposition.

[0079] "Adjusting the programmable gain according to the instantaneous amplitude ratio" refers to dynamically adjusting the amplitude of one signal using a programmable gain amplifier based on the real-time monitored instantaneous amplitude ratio, so that its amplitude reaches a preset proportional relationship or is equal to the amplitude of the other signal. The programmable gain amplifier can change its gain according to the control voltage or current, with the aim of balancing the energy contribution of the two signals and avoiding a situation where one component is too strong or too weak during the superposition process, thereby optimizing the fusion effect.

[0080] "Analog superposition of the output signal of the adjusted first analog processing path with the output signal of the adjusted second analog processing path" refers to the linear superposition of the two signals using an analog adder after phase alignment and amplitude balancing are completed. This superposition process is performed in the analog domain, avoiding quantization errors and delays caused by digital conversion. Its purpose is to seamlessly fuse the signal after high-frequency background purification with the signal after low-frequency physiological feature extraction, forming a complete electrocardiogram signal that contains both high-frequency details and a stable baseline and physiological features.

[0081] The solution in this application continuously monitors the instantaneous phase difference and instantaneous amplitude ratio between the output signals of the first and second analog processing paths, enabling real-time understanding of the dynamic characteristics of the two signals. This real-time monitoring allows the system to precisely adjust the analog delay based on the instantaneous phase difference, ensuring time synchronization between the two signals. Simultaneously, adjusting the programmable gain based on the instantaneous amplitude ratio balances the energy contributions of the two signals. Finally, by analog superimposing the adjusted two signals, effective fusion of high-frequency details and low-frequency physiological characteristics is achieved, avoiding the problems of poor fusion or information loss caused by signal mismatch in traditional methods.

[0082] Through the above technical solution, the accuracy and adaptability of the simulated fusion reconstruction process are significantly improved. Specifically, by monitoring the instantaneous phase difference and instantaneous amplitude ratio in real time and dynamically adjusting the analog delay and programmable gain accordingly, it is possible to ensure optimal matching of signals from different processing paths in terms of time and amplitude, thereby avoiding problems such as phase cancellation and amplitude imbalance that may occur when signals are superimposed. As a result, the obtained fused and reconstructed simulated ECG signal has higher fidelity and can more accurately reflect the true waveform of the original ECG signal, providing high-quality input for subsequent image recognition and analysis, and thus improving the accuracy and reliability of high-frequency ECG waveform analysis.

[0083] In some embodiments described above, continuous monitoring of low-frequency interference characteristics of the original analog ECG signal is proposed. However, if the low-frequency interference characteristics are not analyzed in a refined and multi-dimensional real-time manner, the identification of interference may be inaccurate, thus affecting the dynamic adjustment effect of the bandpass filter bank and limiting the extraction accuracy of low-frequency physiological features. If the above problems are not addressed, in complex physiological signal environments, the system may not be able to effectively filter out various low-frequency noises, thereby reducing the accuracy and reliability of subsequent image recognition and analysis. Therefore, this application further proposes a method for continuously monitoring the low-frequency interference characteristics of the original analog ECG signal. This method continuously monitors the instantaneous energy and instantaneous frequency change rate of the low-frequency component and identifies the start boundary, end boundary, and frequency type of the interference based on the monitoring results, thereby achieving accurate identification and classification of low-frequency interference.

[0084] The aforementioned continuous monitoring of the low-frequency interference characteristics of the raw analog electrocardiogram signal specifically includes: Continuously monitor the instantaneous energy and instantaneous frequency change rate of the low-frequency component of the original simulated electrocardiogram signal; Determine whether the instantaneous energy and the instantaneous frequency change rate exceed a preset threshold; The start and end boundaries of the interference are identified based on the rising and falling edges of the instantaneous energy. The frequency type of the interference is determined based on the instantaneous energy.

[0085] Specifically, instantaneous energy refers to the amount of energy contained in a signal within a very short time window, reflecting the instantaneous strength of the signal. Instantaneous frequency change rate refers to the rate of change of the signal frequency per unit time, characterizing the dynamic characteristics of the signal frequency. By simultaneously monitoring these two parameters, the dynamic characteristics of low-frequency interference can be captured more comprehensively. For example, a sudden increase in instantaneous energy may indicate the presence of interference, while the instantaneous frequency change rate helps distinguish between different types of interference sources. The preset threshold can be understood as a critical value used to distinguish the energy or frequency change rate from normal physiological signals and interference signals. These thresholds can be set based on clinical experience, statistical analysis, or machine learning methods to ensure sensitivity and specificity to interference. When the instantaneous energy or instantaneous frequency change rate exceeds these thresholds, the system can preliminarily determine the presence of low-frequency interference.

[0086] In practical applications, the rising edge of instantaneous energy refers to the moment when energy rapidly rises from a lower level to a higher level, and the falling edge refers to the moment when energy rapidly falls from a higher level to a lower level. By identifying these edges, the start and end times of interference can be accurately determined, thus providing a precise time window for subsequent filter parameter adjustments. Furthermore, determining the frequency type of interference based on instantaneous energy means analyzing the characteristics of the instantaneous energy, such as its duration, amplitude, and correlation with a specific frequency range, to determine the nature of the interference. For example, low-frequency energy with a long duration and slowly changing amplitude may indicate baseline drift, while periodically occurring energy with a large amplitude may indicate power frequency interference or motion artifacts.

[0087] This application's solution addresses the potential accuracy limitations of traditional monitoring methods by continuously monitoring the instantaneous energy and instantaneous frequency change rate of the low-frequency component of the original simulated electrocardiogram (ECG) signal. Real-time tracking of these two key parameters enables the system to more sensitively detect the occurrence and changes of low-frequency interference. By determining whether these parameters exceed preset thresholds, normal physiological activity can be quickly distinguished from abnormal interference. Furthermore, the rising and falling edges of the instantaneous energy are used to accurately identify the start and end boundaries of interference, ensuring accurate control of the interference's time window. Moreover, the frequency type of interference is determined based on the characteristics of the instantaneous energy, allowing the system to classify and process low-frequency interference of different natures. This refined interference identification and classification capability enables more accurate and timely dynamic adjustment of the filtering parameters of the subsequent bandpass filter bank, effectively suppressing various types of low-frequency interference and ensuring the purity of low-frequency physiological characteristics.

[0088] Through the above technical solution, this application enables high-precision, multi-dimensional real-time monitoring of low-frequency interference characteristics in raw analog electrocardiogram (ECG) signals. Compared to traditional methods that rely on a single indicator or make rough judgments, this application significantly improves the ability to identify and respond to different types of low-frequency interference by combining instantaneous energy and instantaneous frequency change rate. Therefore, the system can more accurately identify the occurrence, duration, and specific type of interference, providing a more reliable and refined basis for the dynamic adjustment of the bandpass filter bank. This precise interference identification and classification capability effectively avoids the problems of over-filtering or under-filtering, thereby ensuring the accuracy and completeness of low-frequency physiological feature extraction and greatly improving the overall performance and reliability of the image recognition-based high-frequency ECG waveform analysis method.

[0089] In some preferred embodiments, a specific example is given below. Suppose that during monitoring, the low-frequency portion of the original analog electrocardiogram signal exhibits the following characteristics: First, slight patient movement can cause a slow baseline drift. In this case, the continuously monitored instantaneous energy will show a slow upward or downward trend, but the instantaneous frequency change rate is relatively small. The system will determine that the instantaneous energy exceeds a lower threshold set for baseline drift and identify the start and end times of the drift based on the rising and falling edges of the energy. Based on the slow change characteristics of the instantaneous energy, it is determined to be an interference of the baseline drift type.

[0090] Secondly, when power frequency interference is present (e.g., 50Hz or 60Hz power supply noise generating harmonics or modulation effects in the low-frequency range), the continuously monitored instantaneous energy may periodically fluctuate with high amplitude, and the instantaneous frequency change rate exhibits regular changes around specific frequency points. The system determines that both the instantaneous energy and the instantaneous frequency change rate exceed a higher threshold set for power frequency interference, and identifies it as power frequency interference based on its periodic characteristics and energy amplitude. In this way, the system can accurately distinguish different types of low-frequency interference, such as baseline drift and power frequency interference, and identify their precise occurrence periods based on the combined characteristics of instantaneous energy and instantaneous frequency change rate. Consequently, the subsequent bandpass filter bank can dynamically adjust its filtering parameters, such as adjusting the cutoff frequency, bandwidth, or gain, based on the identified interference type and time window, to specifically suppress specific interference while preserving useful low-frequency physiological characteristics to the maximum extent.

[0091] Specifically, the continuous monitoring of the instantaneous energy of the low-frequency component of the original analog electrocardiogram signal can be achieved in the following manner.

[0092] The continuous monitoring of the instantaneous energy of the low-frequency component of the original simulated electrocardiogram signal includes: The low-frequency component of the original analog electrocardiogram signal is input into multiple parallel analog square-law detectors, so that each analog square-law detector performs energy integration on the low-frequency component of the original analog electrocardiogram signal according to a different time constant, resulting in multiple integration results. The instantaneous energy of the low-frequency component is obtained by dynamically weighting and averaging the multiple integral results.

[0093] Specifically, multiple parallel analog square-law detectors are configured to simultaneously process the low-frequency portion of the raw analog ECG signal. Each analog square-law detector is designed with a different time constant, meaning they have different speeds of response and smoothness to the energy integral of the input signal.

[0094] For example, one detector might use a shorter time constant to respond quickly to instantaneous energy changes, while another detector might use a longer time constant to provide a smoother, more stable energy trend. In this way, instantaneous energy information in the low-frequency range can be captured at different time scales. Here, energy integration refers to accumulating the instantaneous power of the input signal to reflect its energy content within a certain time window. The integral result output by each analog square-law detector represents a measurement of the low-frequency energy at a specific time constant.

[0095] Furthermore, a dynamic weighted average is applied to the multiple integration results to synthesize the outputs of detectors with different time constants, thereby obtaining a more robust and accurate instantaneous energy estimate. The dynamic weighted average can adjust the weights of each integration result based on the real-time characteristics of the signal (e.g., signal stability, noise level, or the presence of specific interference). For example, when the signal changes drastically, a higher weight can be assigned to the short-time-constant detector to capture rapid changes; when the signal is relatively stable, a higher weight can be assigned to the long-time-constant detector to reduce the impact of noise. Thus, the instantaneous energy of the low-frequency component can be obtained, which more comprehensively and accurately reflects the actual situation of low-frequency interference.

[0096] The proposed solution inputs the low-frequency component of the original analog electrocardiogram signal into multiple parallel analog square-law detectors, with each detector integrating energy according to a different time constant. This allows for the capture of instantaneous energy information of low-frequency interference across multiple time scales. This multi-time-scale analysis avoids the problems of underestimation or over-smoothing of rapidly or slowly changing energy that can occur with a single time constant. Furthermore, by dynamically weighting and averaging the integration results under these different time constants, the contribution of each detector output can be adaptively adjusted according to the real-time characteristics of the signal. This ensures a rapid response to instantaneous energy while effectively suppressing the effects of noise and artifacts, resulting in more accurate and reliable monitoring of the instantaneous energy of the low-frequency component.

[0097] The above technical solution overcomes the limitations of single energy detection methods when facing complex and variable low-frequency interference. Specifically, by combining multiple parallel analog square-law detectors with different time constants, the system can simultaneously achieve rapid response to low-frequency interference and smooth suppression of background noise. The introduction of dynamic weighted averaging further enhances the system's adaptability to low-frequency interference of different types and intensities, ensuring high-precision and robust low-frequency instantaneous energy monitoring results under various physiological and environmental conditions. This provides a solid data foundation for accurately identifying the start and end boundaries of interference and determining the interference frequency type, significantly improving the accuracy and reliability of overall ECG signal analysis.

[0098] In some embodiments described above, continuous monitoring of the instantaneous frequency change rate of the low-frequency component of the original simulated ECG signal is proposed. However, in practical applications, the low-frequency component of the original simulated ECG signal may contain multiple independent and dynamically changing interference sources, such as respiratory artifacts, electromyography artifacts, or baseline drift. The instantaneous frequencies of these interference sources may cross, merge, or change rapidly, making it difficult for a single or statically configured frequency tracking mechanism to accurately and continuously identify and track the instantaneous frequency and its rate of change of each independent interference source. If the above problems are not addressed, it may lead to misjudgment of the low-frequency interference characteristics, thereby affecting the adaptive adjustment effect of subsequent simulation parameters and reducing the accuracy of ECG signal pre-correction.

[0099] In response, this application further proposes a method for continuously monitoring the instantaneous frequency change rate of the low-frequency component of a raw analog electrocardiogram signal, comprising: The low-frequency portion of the original analog electrocardiogram signal is input into multiple parallel analog frequency tracking units, so that each analog frequency tracking unit has a preset capture range and tracking bandwidth for a specific frequency. Real-time tracking of instantaneous frequencies within the capture range, and outputting the corresponding instantaneous frequency signal; When the instantaneous frequency signals of multiple analog frequency tracking units cross or merge, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted according to the signal strength and frequency change trend of each analog frequency tracking unit to ensure that the instantaneous frequency of each independent interference source is continuously tracked. The instantaneous frequency signal output by each of the analog frequency tracking units is processed by analog differentiation to obtain the instantaneous frequency change rate of each independent interference source.

[0100] Specifically, the aforementioned method for continuously monitoring the instantaneous frequency change rate of the low-frequency component of the original analog ECG signal aims to improve the tracking accuracy and robustness against complex low-frequency interference signals through parallel processing and dynamic adaptive mechanisms. Here, "multiple parallel analog frequency tracking units" refers to multiple independent analog circuit modules, each designed to track signals within a specific frequency range. These units operate in parallel, enabling them to simultaneously process multiple potential interference components from the low-frequency portion of the original analog ECG signal, avoiding the limitations of a single tracker in the face of multiple interference sources. Each analog frequency tracking unit is "preset with a specific frequency capture range and tracking bandwidth," meaning that during design or configuration, an initial frequency capture range and tracking sensitivity are set based on the expected type of low-frequency interference (e.g., respiratory rate range, electromyographic interference frequency range) to allow it to focus on specific types of interference.

[0101] In practical applications, each analog frequency tracking unit "tracks the instantaneous frequency within the capture range in real time and outputs the corresponding instantaneous frequency signal." This is typically achieved through a phase-locked loop (PLL), a frequency discriminator, or other analog frequency tracking circuits, with the aim of continuously outputting the currently tracked frequency value. When "the instantaneous frequency signals of multiple analog frequency tracking units intersect or merge," for example, when the frequencies of two different interference sources become very close or temporarily overlap at a certain moment, the system "dynamically adjusts the tracking bandwidth and center frequency of each analog frequency tracking unit according to the signal strength and frequency change trend of each analog frequency tracking unit." This means that the system can intelligently identify such complex situations and adjust its tracking parameters according to the signal energy magnitude and frequency change direction of each tracking unit (e.g., whether it is rising or falling).

[0102] For example, for interference sources with weak signal strength but obvious frequency change trends, their tracking bandwidth may be appropriately increased to prevent loss of lock; for interference sources with strong signal strength and stable frequency, a narrower bandwidth may be maintained to improve accuracy. The purpose of this dynamic adjustment is to "ensure that the instantaneous frequency of each independent interference source is continuously tracked," maintaining independent identification and tracking of each interference component even in complex and intertwined interference signals. Finally, "analog differentiation processing is performed on the instantaneous frequency signal output by each of the analog frequency tracking units," directly calculating the rate of change of the instantaneous frequency over time through an analog differentiating circuit, thereby obtaining the instantaneous frequency change rate of each independent interference source.

[0103] This application's solution introduces multiple parallel analog frequency tracking units, enabling the system to simultaneously monitor and process multiple potential low-frequency interference sources, overcoming the limitations of traditional single trackers in complex interference environments. When the instantaneous frequency signals of interference sources cross or merge, a dynamic adjustment mechanism based on signal strength and frequency change trends intelligently optimizes the parameters of each tracking unit, effectively avoiding tracking loss or misjudgment and ensuring continuous tracking of each independent interference source. Therefore, by performing analog differential processing on the instantaneous frequency signals output by each tracking unit, the instantaneous frequency change rate of each independent interference source can be directly and in real-time obtained, providing accurate and multi-dimensional information for subsequent low-frequency interference characteristic analysis and adaptive filtering.

[0104] The above technical solution significantly improves the monitoring accuracy and robustness of the instantaneous frequency change rate of complex interference sources in the low-frequency portion of the original analog ECG signal. Compared to methods that only perform simple frequency monitoring, this application can effectively handle the frequency crossover or fusion of multiple interference sources, ensuring that the frequency change trend of each independent interference source is accurately captured. This allows the system to more precisely identify the dynamic characteristics of low-frequency interference, thus providing a more reliable basis for the subsequent dynamic adjustment of the bandpass filter bank's filtering parameters, ultimately improving the overall performance of analog domain pre-calibration processing and further ensuring the accuracy of high-frequency ECG waveform analysis.

[0105] As a specific implementation method, a concrete example is given below. Suppose the low-frequency portion of the original simulated ECG signal contains two main types of interference: a respiratory artifact with a frequency between 0.1 Hz and 0.5 Hz, and a slight electromyography (EMG) artifact with a frequency between 0.5 Hz and 2 Hz. To accurately track the instantaneous frequency change rate of these two interferences, two parallel simulated frequency tracking units can be configured. The first simulated frequency tracking unit is preset to capture a range of 0.05 Hz to 0.6 Hz with a tracking bandwidth of 0.1 Hz, primarily used to track the respiratory artifact. The second simulated frequency tracking unit is preset to capture a range of 0.4 Hz to 2.5 Hz with a tracking bandwidth of 0.2 Hz, primarily used to track the EMG artifact. Under normal circumstances, these two units independently track their target frequencies. However, when the patient takes a deep breath or engages in slight activity, the respiratory rate may temporarily increase, approaching or crossing the low-end frequency of the EMG artifact. At this time, the system continuously monitors the instantaneous frequency signals output by both simulated frequency tracking units.

[0106] For example, if the system detects a high signal strength and rising frequency from the first analog frequency tracking unit, while the second analog frequency tracking unit has a low signal strength and slightly fluctuating frequency, it dynamically adjusts their tracking parameters based on this information. Specifically, it might temporarily increase the tracking bandwidth of the first analog frequency tracking unit to ensure it can continuously capture rapidly changing respiratory rates, while fine-tuning the center frequency of the second analog frequency tracking unit to prevent it from being "pulled off course" by respiratory artifact signals. Through this dynamic adjustment, even in complex situations of frequency crossover or fusion, two independent interference sources can be continuously and accurately tracked. Ultimately, the instantaneous frequency signal output by each analog frequency tracking unit is fed into its respective analog differentiating circuit to calculate the instantaneous frequency change rate of both respiratory and electromyographic artifacts in real time.

[0107] In some embodiments described above in this application, when the instantaneous frequency signals of multiple analog frequency tracking units cross or merge, it is necessary to dynamically adjust their tracking bandwidth and center frequency according to the signal strength and frequency change trend of each analog frequency tracking unit to ensure that the instantaneous frequency of each independent interference source is continuously tracked. However, in practical applications, adjusting solely based on simple signal strength and frequency change trends may lead to insufficient adjustment accuracy or even misjudgment when the signal strength of the interference sources differs significantly or the frequency change trend is complex, thereby affecting the accurate differentiation and continuous tracking of independent interference sources.

[0108] In response, this application further proposes the following steps for dynamically adjusting the tracking bandwidth and center frequency of each analog frequency tracking unit based on the signal strength and frequency change trend of each unit when the instantaneous frequency signals of multiple analog frequency tracking units cross or merge, so as to ensure that the instantaneous frequency of each independent interference source is continuously tracked: The signal strength of each analog frequency tracking unit is normalized. Based on the normalized signal strength and the frequency change trend, a weighting factor is dynamically assigned to each analog frequency tracking unit. Based on the weighting factor, increase the weight of the analog frequency tracking unit with low signal strength; Based on the normalized signal strength and the frequency change trend, the instantaneous frequency of each analog frequency tracking unit is compared in real time through an analog comparator array. Identify frequency crossover or fusion points; At the frequency crossover or fusion point, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted according to the weighted signal strength of the analog frequency tracking unit and the frequency change trend.

[0109] Specifically, the signal strength of each analog frequency tracking unit is normalized to eliminate the impact of signal strength differences between different analog frequency tracking units, making subsequent weight allocation and comparison more fair and accurate. For example, the signal strength can be adjusted to a uniform range through logarithmic compression, differential processing, and adjusting the gain of a programmable gain amplifier.

[0110] The process involves dynamically assigning weight factors to each analog frequency tracking unit (FRTU) based on the normalized signal strength and frequency variation trends. This aims to assign different levels of importance to different FRTUs according to the actual characteristics of the interference source. For example, an interference source with low signal strength but a significant frequency variation trend can be assigned a higher weight to ensure it is not masked by strong signal interference. In practical applications, increasing the weight of FRTUs with low signal strength based on the weight factor is to prevent weak signal interference sources from being ignored when there are significant differences in signal strength among multiple interference sources. By increasing their weight, even if their original signal strength is low, they can receive sufficient consideration in subsequent adjustment decisions.

[0111] Furthermore, based on the normalized signal strength and frequency variation trends, the instantaneous frequencies of each analog frequency tracking unit are compared in real time using an analog comparator array. This aims to provide high-precision real-time frequency comparison capabilities. The analog comparator array can quickly and accurately detect the proximity, intersection, or merging states between frequencies tracked by different tracking units. Therefore, identifying frequency intersection or merging points refers to determining the critical moment when different interference source frequencies interact based on the output of the analog comparator array. Further, at frequency intersection or merging points, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted based on the weighted signal strength and frequency variation trends of the analog frequency tracking units. This aims to make precise adjustments using more comprehensive information (weighted signal strength) at the most critical moment, thereby effectively distinguishing and tracking interference sources that are close to or overlapping each other.

[0112] This application's solution addresses the problem of insufficient tracking accuracy caused by signal strength differences or complex frequency changes when instantaneous frequency signals from multiple analog frequency tracking units intersect or merge. By introducing signal strength normalization processing and a dynamic weight allocation mechanism, it solves this problem. Through this technical solution, the application significantly improves the tracking accuracy and robustness of low-frequency interference in high-frequency ECG waveforms under multi-interference-source environments. Especially in scenarios with large differences in interference source signal strength and complex frequency changes, normalization processing and dynamic weight allocation effectively prevent weak signal interference sources from being masked or misjudged, ensuring continuous tracking of all independent interference sources. Therefore, this application provides more accurate low-frequency physiological feature extraction, offering higher-quality pre-correction signals for subsequent image recognition and analysis, thereby improving the reliability and diagnostic accuracy of overall ECG waveform analysis.

[0113] In some preferred embodiments, it is assumed that there are two low-frequency interference sources whose instantaneous frequency signals intersect at a certain moment. Interference source A has a higher signal strength, while interference source B has a relatively lower signal strength. First, the signal strengths tracked by analog frequency tracking units A and B are normalized to ensure they are compared on a uniform scale. Then, based on the normalized signal strength and frequency change trends, weighting factors are dynamically assigned to analog frequency tracking units A and B. Because interference source B has a lower signal strength, its corresponding weighting factor is appropriately increased to ensure it receives sufficient attention in subsequent decisions.

[0114] Subsequently, the instantaneous frequencies of analog frequency tracking units A and B are compared in real time using an analog comparator array to accurately identify the moment of their crossover. Upon identifying the frequency crossover point, the system dynamically adjusts the tracking bandwidth and center frequency of analog frequency tracking units A and B based on their weighted signal strengths (with the weight of analog frequency tracking unit B being increased) and their respective frequency change trends. For example, the tracking bandwidth of analog frequency tracking unit A may be appropriately narrowed to more accurately lock its frequency; while the center frequency of analog frequency tracking unit B may be fine-tuned to avoid frequency confusion with that of analog frequency tracking unit A, thereby ensuring that the instantaneous frequencies of the two independent interference sources can be continuously and accurately tracked, even if they are very close in frequency.

[0115] Specifically, the above-mentioned normalization process for the signal strength of each analog frequency tracking unit may include the following steps: Logarithmically compress the signal strength of each analog frequency tracking unit to obtain a logarithmically compressed signal; The logarithmically compressed signal is differentially processed to obtain a differential signal; The gain of the programmable gain amplifier is dynamically adjusted based on the instantaneous amplitude of the differential signal. The differential signal after gain adjustment is normalized.

[0116] The signal strength of each analog frequency tracking unit is logarithmically compressed to convert the wide dynamic range signal strength into a more easily processed logarithmic domain signal. This effectively suppresses the impact of drastic signal strength fluctuations on subsequent processing and highlights the relative changes in the signal. Specifically, this function can be achieved using an analog logarithmic amplifier, whose output voltage is proportional to the logarithm of the input signal strength.

[0117] Furthermore, differential processing is applied to the logarithmically compressed signal to extract instantaneous information about signal strength changes and eliminate DC or slowly varying components in the signal. This allows for more sensitive detection of rapidly changing signal strength trends, which is crucial for identifying the start, end, and frequency crossover or merging points of interference sources. In practical applications, this can be achieved using analog differentiating circuits, such as operational amplifiers and RC networks.

[0118] Furthermore, the gain of the programmable gain amplifier is dynamically adjusted based on the instantaneous amplitude of the differential signal to ensure that the amplitude of the signal after differential processing is always within an optimal dynamic range, avoiding signal overload or excessively low signal-to-noise ratio. Specifically, a feedback control loop can be designed to compare the instantaneous amplitude of the differential signal with a preset threshold and adjust the gain of the programmable gain amplifier in real time to optimize the quantization accuracy and processing effect of the signal.

[0119] Finally, the differential signal after gain adjustment is normalized to unify the signal strength to a standardized range (e.g., 0 to 1 or -1 to 1) to facilitate fair comparison of signal strength between different analog frequency tracking units and accurate allocation of subsequent weighting factors. This can be achieved using an analog divider or scaling circuit.

[0120] This application's solution addresses the problem of inaccurate weight allocation caused by excessively large differences in signal strength when the instantaneous frequency signals of multiple analog frequency tracking units intersect or merge. By performing refined normalization on the signal strength of the analog frequency tracking units, this application provides a more robust and accurate method for normalizing the signal strength of analog frequency tracking units. This method effectively overcomes the inaccuracies or response delays that traditional normalization methods may encounter when handling wide dynamic ranges and rapidly changing signal strengths. Therefore, in complex scenarios where multiple interference sources coexist and frequencies intersect or merge, the signal strength of each analog frequency tracking unit can be more accurately evaluated, leading to more refined and reliable dynamic adjustments to the tracking bandwidth and center frequency. This significantly improves the continuous tracking capability against independent interference sources and the overall system's anti-interference performance.

[0121] refer to Figure 4 , Figure 4 This is a structural diagram of a high-frequency electrocardiogram waveform analysis system based on image recognition provided in an embodiment of the present invention, comprising: The detection end is used to receive the raw analog ECG signal; the raw analog ECG signal is subjected to analog domain pre-correction processing, the analog domain pre-correction processing includes: real-time monitoring of the signal characteristics of the raw analog ECG signal, the signal characteristics including baseline position and intensity of rapidly changing components; The adjustment terminal is used to adaptively adjust the analog parameters of the original simulated electrocardiogram signal according to the signal characteristics, the analog parameters including baseline position and rapidly changing component gain; The recognition end is used to apply the simulated electrocardiogram signal, which has undergone analog domain pre-correction processing, to image recognition and analysis.

[0122] This system aims to address the signal quality degradation problem caused by the lack of adaptive pre-correction capability in traditional high-frequency electrocardiogram (ECG) waveform analysis methods when processing raw ECG signals from diverse devices. By setting up a detection end, an adjustment end, and an identification end, the system of this application can perform real-time, adaptive pre-correction processing on raw ECG signals in the analog domain.

[0123] Specifically, the detection end receives the raw analog ECG signal and monitors its signal characteristics in real time, such as baseline position and the intensity of rapidly changing components. The adjustment end adaptively adjusts the analog parameters of the raw analog ECG signal, including baseline position and the gain of rapidly changing components, based on these monitored signal characteristics, effectively eliminating interference and optimizing signal quality. Finally, the recognition end uses the pre-corrected, high-quality analog ECG signal for image recognition analysis, significantly improving the accuracy and reliability of diagnosis. This systematic design ensures quality control throughout the entire chain from signal acquisition to image recognition, making it particularly suitable for complex and variable real-world application scenarios such as primary healthcare institutions.

[0124] To better understand the technical solution proposed in this application, the various components of the system are described in detail below.

[0125] The specific methods and objectives for receiving the original analog ECG signal, analog domain pre-calibration processing, signal characteristic monitoring, analog parameter adjustment, and image recognition analysis have been described in the above embodiments, and will not be repeated here. It should be emphasized that this application modularizes these functions into specific system components to achieve more efficient and stable operation.

[0126] Specifically, the detection end can be understood as the hardware or software module in the system responsible for the input and preliminary processing of the raw analog ECG signal. For example, the detection end may include an analog signal input interface for connecting to the ECG acquisition device, and an analog signal processing unit that integrates circuitry for real-time monitoring of the signal characteristics of the raw analog ECG signal (including baseline position and intensity of rapidly changing components). This analog signal processing unit may consist of a series of analog sensors, comparators, and filters for continuously acquiring and analyzing the instantaneous state of the ECG signal.

[0127] In one implementation, the detection end can be a standalone hardware module integrating an analog front-end amplifier, low-pass filter, high-pass filter, baseline drift detection circuit, and high-frequency component intensity detection circuit. These circuits work together to output the baseline position information and rapidly changing component intensity information of the signal in real time. In another implementation, the detection end can be a programmable analog front-end (AFE) chip, which performs signal reception and characteristic monitoring functions by configuring its internal registers.

[0128] The adjustment unit is the core component of the system that adaptively adjusts the analog parameters of the original analog ECG signal based on the signal characteristics output from the detection unit. For example, the adjustment unit can be an analog feedback control circuit that receives baseline position and rapidly changing component intensity information from the detection unit and outputs corresponding control signals to adjust the analog parameters according to a preset control algorithm. These analog parameters include the baseline position and the gain of the rapidly changing component. In one specific implementation, the adjustment unit can include a programmable DC bias circuit and a programmable gain amplifier. When the detection unit reports baseline drift, the adjustment unit restores the ECG signal baseline to the reference potential by adjusting the output of the DC bias circuit. When the intensity of the rapidly changing component needs adjustment, the adjustment unit controls the gain of the programmable gain amplifier to ensure that key details in the high-frequency QRS waveform are appropriately amplified or attenuated. The adjustment unit can also be a combination of a microcontroller (MCU) and a digital-to-analog converter (DAC). The MCU calculates the required analog parameter adjustment amount based on the detected signal characteristics and outputs an analog control voltage through the DAC to drive the analog circuit for parameter adjustment.

[0129] The recognition unit is the module in the system that uses the analog ECG signal, after analog domain pre-calibration, for image recognition and analysis. For example, the recognition unit may include a high-performance analog-to-digital converter (ADC) to convert the pre-calibrated analog signal into a digital signal, and a digital signal processor (DSP) or embedded system to convert the digital ECG signal into an image format. This image format can be a grayscale image or a color image, where the morphology, amplitude, and timing information of the ECG waveform are encoded as the brightness or color of the image pixels. As one implementation, the recognition unit can be a dedicated image processing module integrating an ADC, an image processing unit, and a communication interface. This module sends the processed ECG waveform image to an external image recognition server or a local machine learning inference engine for analysis via the communication interface. In another implementation, the recognition unit can directly integrate a lightweight convolutional neural network (CNN) inference accelerator to achieve localized image feature extraction and preliminary recognition, thereby reducing data transmission latency and improving real-time performance.

[0130] Compared with existing technologies, the high-frequency electrocardiogram waveform analysis system based on image recognition proposed in this application has the core innovation of integrating adaptive pre-correction processing in the analog domain into the system architecture. Through the collaborative work of dedicated detection, adjustment, and recognition ends, it fundamentally solves the problem of inconsistent quality of raw electrocardiogram signals. Traditional systems typically process signals after digitization, and their fixed processing logic struggles to effectively handle the complex interference and equipment differences present in analog signals during the acquisition stage, resulting in limited image recognition accuracy.

[0131] The system in this application monitors signal characteristics in real time at the detection end, and the adjustment end adaptively adjusts analog parameters based on these characteristics. This ensures that the ECG signal baseline is stable, high-frequency components are clear, and the signal-to-noise ratio is significantly improved before the signal is digitized. This systematic "correction before recognition" strategy enables the recognition end to perform image conversion and recognition analysis based on higher-quality analog signals, thereby significantly improving the accuracy and reliability of diagnosis. For example, in primary healthcare institutions facing baseline drift or high-frequency component attenuation introduced by aging equipment, this system can effectively suppress these interferences through adaptive adjustment of the analog domain, avoiding artifacts and information loss caused by fixed processing logic in traditional systems, and significantly improving the system's robustness and diagnostic efficiency in complex real-world application environments.

[0132] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A high-frequency electrocardiogram waveform analysis method based on image recognition, characterized in that, include: Receive raw analog electrocardiogram signals; The original simulated electrocardiogram (ECG) signal is subjected to analog domain pre-correction processing, which includes: real-time monitoring of the signal characteristics of the original simulated ECG signal, including baseline position and intensity of rapidly changing components; Based on the signal characteristics, the analog parameters of the original simulated electrocardiogram signal are adaptively adjusted, including the baseline position and the gain of the rapidly changing component; The simulated electrocardiogram signal, after being pre-corrected in the analog domain, is used for image recognition and analysis.

2. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 1, characterized in that, The analog parameters of the original simulated electrocardiogram signal are adaptively adjusted according to the signal characteristics. These analog parameters include baseline position and gain of rapidly changing components, including: The original simulated electrocardiogram signal is split into a first simulated processing path and a second simulated processing path; According to the first simulation processing path, a variable cutoff frequency high-pass filter is used to perform high-frequency background purification on the original simulated electrocardiogram signal. According to the second simulation processing path, a bandpass filter bank is used to extract low-frequency physiological features from the original simulated electrocardiogram signal. The output signals of the first and second analog processing paths are simulated and fused together to reconstruct the original analog ECG signal, thereby adaptively adjusting the analog parameters of the original analog ECG signal, including baseline position and gain of rapidly changing components.

3. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 2, characterized in that, The step of extracting low-frequency physiological features from the original simulated electrocardiogram signal using a bandpass filter bank according to the second simulation processing path includes: Spectral analysis was performed on the original simulated electrocardiogram signal to determine the baseline drift interference characteristics in the low-frequency band; Based on the baseline drift interference characteristics, the cutoff frequency or quality factor of at least one filter in the bandpass filter bank is dynamically adjusted to suppress the baseline drift interference. For the signal processed by the bandpass filter bank, the signal components within the preset low-frequency band are extracted as the low-frequency physiological features; The passband gain of the bandpass filter bank is adaptively fine-tuned based on the amplitude changes of the low-frequency physiological characteristics.

4. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 2, characterized in that, The step of performing analog fusion reconstruction of the output signal of the first analog processing path and the output signal of the second analog processing path includes: Continuously monitor the instantaneous phase difference and instantaneous amplitude ratio between the output signal of the first analog processing path and the output signal of the second analog processing path; Adjust the simulation delay based on the instantaneous phase difference; Adjust the programmable gain according to the instantaneous amplitude ratio; The output signal of the adjusted first analog processing path is superimposed with the output signal of the adjusted second analog processing path.

5. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 3, characterized in that, The continuous monitoring of the low-frequency interference characteristics of the original simulated electrocardiogram signal includes: Continuously monitor the instantaneous energy and instantaneous frequency change rate of the low-frequency component of the original simulated electrocardiogram signal; Determine whether the instantaneous energy and the instantaneous frequency change rate exceed a preset threshold; The start and end boundaries of the interference are identified based on the rising and falling edges of the instantaneous energy. The frequency type of the interference is determined based on the instantaneous energy.

6. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 5, characterized in that, The continuous monitoring of the instantaneous energy of the low-frequency component of the original simulated electrocardiogram signal includes: The low-frequency component of the original analog electrocardiogram signal is input into multiple parallel analog square-law detectors, so that each analog square-law detector performs energy integration on the low-frequency component of the original analog electrocardiogram signal according to a different time constant, resulting in multiple integration results. The instantaneous energy of the low-frequency component is obtained by dynamically weighting and averaging the multiple integral results.

7. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 5, characterized in that, The continuous monitoring of the instantaneous frequency change rate of the low-frequency component of the original simulated electrocardiogram signal includes: The low-frequency portion of the original analog electrocardiogram signal is input into multiple parallel analog frequency tracking units, so that each analog frequency tracking unit has a preset capture range and tracking bandwidth for a specific frequency. Real-time tracking of instantaneous frequencies within the capture range, and outputting the corresponding instantaneous frequency signal; When the instantaneous frequency signals of multiple analog frequency tracking units cross or merge, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted according to the signal strength and frequency change trend of each analog frequency tracking unit to ensure that the instantaneous frequency of each independent interference source is continuously tracked. The instantaneous frequency signal output by each of the analog frequency tracking units is processed by analog differentiation to obtain the instantaneous frequency change rate of each independent interference source.

8. The high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 7, characterized in that, When the instantaneous frequency signals of multiple analog frequency tracking units cross or merge, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted according to the signal strength and frequency change trend of each unit to ensure that the instantaneous frequency of each independent interference source is continuously tracked, including: The signal strength of each of the analog frequency tracking units is normalized. Based on the normalized signal strength and the frequency change trend, a weighting factor is dynamically assigned to each of the analog frequency tracking units. Based on the weighting factor, increase the weight of the analog frequency tracking unit with low signal strength; Based on the normalized signal strength and the frequency change trend, the instantaneous frequency of each analog frequency tracking unit is compared in real time through an analog comparator array. Identify frequency crossover or fusion points; At the frequency crossover or fusion point, the tracking bandwidth and center frequency of each analog frequency tracking unit are dynamically adjusted according to the weighted signal strength of the analog frequency tracking unit and the frequency change trend.

9. A high-frequency electrocardiogram waveform analysis method based on image recognition according to claim 8, characterized in that, The normalization process for the signal strength of each of the analog frequency tracking units includes: Logarithmically compress the signal strength of each of the analog frequency tracking units to obtain a logarithmically compressed signal; The logarithmically compressed signal is differentially processed to obtain a differential signal; The gain of the programmable gain amplifier is dynamically adjusted based on the instantaneous amplitude of the differential signal. The differential signal after gain adjustment is normalized.

10. A high-frequency electrocardiogram waveform analysis system based on image recognition, characterized in that, include: The detection end is used to receive raw analog electrocardiogram signals; The original simulated electrocardiogram (ECG) signal is subjected to analog domain pre-correction processing, which includes: real-time monitoring of the signal characteristics of the original simulated ECG signal, including baseline position and intensity of rapidly changing components; The adjustment terminal is used to adaptively adjust the analog parameters of the original simulated electrocardiogram signal according to the signal characteristics, the analog parameters including baseline position and rapidly changing component gain; The recognition end is used to apply the simulated electrocardiogram signal, which has undergone analog domain pre-correction processing, to image recognition and analysis.

Images