An arrhythmia diagnosis system based on multi-sensor fusion

Abstract

The present invention relates to the field of intelligent medical technology, and in particular to an arrhythmia diagnosis system based on multi-sensor fusion, the system comprising: an electrocardiogram (ECG) signal module, comprising an ECG signal sensor for collecting ECG signals; a pulse wave signal module, comprising a pulse wave sensor for collecting pulse wave signals; a three-axis acceleration module, comprising an acceleration sensor for collecting acceleration signals; a main control chip module, serving as a data processing module, for receiving and analyzing data to obtain an arrhythmia diagnosis result; a Bluetooth module, serving as a communication module, for transmitting the ECG signal, the pulse wave signal, the acceleration signal, and the arrhythmia diagnosis result to a mobile terminal; and a power management module, comprising a charging, power supply, and switching circuit for providing power supply for the system. The system provided by the present invention is stable, portable, and has low power consumption, and is of great significance for the prevention of cardiovascular diseases and ECG monitoring via the Internet of Things.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and in particular to an arrhythmia diagnosis system based on multi-sensor fusion. Background Art

[0002] Heart disease is a disease that poses a great threat to human health. Electrocardiogram, as a means of examination to record the electrical activity of the heart, can reflect the complete process of the heart. A 12-lead system is often used clinically, but traditional 12-lead devices have portability issues. Wearable electrocardiogram devices have become a research hotspot. With the rapid development of wearable computing technology, it has had a significant impact on the health and medical fields. Wearable devices are small and light, suitable for daily long-term monitoring. However, wearable devices still need to be optimized in terms of accuracy and reliability, and sensor technology needs to be improved to meet a wider range of application needs. The present invention combines electronic information technology, communication technology, biomedicine and other technical theories to disclose a dynamic electrocardiogram monitoring terminal for arrhythmia diagnosis, which uses three different sensor modules, an electrocardiogram signal sensor, a pulse wave sensor and an acceleration sensor, to perform dynamic electrocardiogram monitoring on the subject.

[0003] At present, research on ECG signal filtering algorithms has achieved certain results; however, during use, wearable ECG monitoring devices often produce baseline drift and motion artifact interference due to human activity, especially the ECG signal frequency during exercise is similar to the interference frequency, making it difficult to separate; traditional adaptive filtering methods perform poorly when processing non-stationary noise and rapidly changing signals, and signal processing algorithms need to be improved; the present invention discloses an optimized filtering algorithm that processes the collected information, effectively removes the electromyographic signal and motion artifacts in the ECG signal, obtains a cleaner ECG signal, and improves the robustness and adaptability in complex noisy environments.

[0004] Traditional electrocardiogram analysis relies on manual labeling by experts, which is time-consuming, labor-intensive, and easily influenced by subjective factors. In recent years, deep learning-based methods can automatically extract electrocardiogram features to achieve efficient and accurate arrhythmia classification, but the increased model complexity also brings challenges to computing resources. To this end, the present invention discloses an arrhythmia classification model that introduces dynamic convolution and multi-scale feature fusion strategies to reduce model complexity. It maintains efficient arrhythmia classification and reduces computing resource requirements. Users can diagnose in real time, and data is synchronized to mobile terminals, facilitating remote monitoring by professional medical staff and family members, providing a convenient, low-cost, and comfortable technical solution for preventing heart health risks. Summary of the Invention

[0005] In response to the defects in the prior art and in combination with the needs of practical applications, the present invention provides an arrhythmia diagnosis system based on multi-sensor fusion, the system comprising: an electrocardiogram signal module, the electrocardiogram signal module comprising an electrocardiogram signal sensor, for collecting the electrocardiogram signal of a subject; a pulse wave signal module, the pulse wave signal module comprising a pulse wave sensor, for collecting the pulse wave signal of the subject; a three-axis acceleration module, the three-axis acceleration module comprising an acceleration sensor, for collecting the acceleration signal of the subject; a main control chip module, the main control chip module serving as a data processing module of the system, for data reception and data analysis, so as to obtain an arrhythmia diagnosis result; a Bluetooth module, the Bluetooth module serving as a communication module of the system, for transmitting the electrocardiogram signal, the pulse wave signal, the acceleration signal and the arrhythmia diagnosis result to a mobile terminal; a power management module, the power management module comprising charging, power supply and switching circuits, for providing power supply for the system. The present invention integrates electrocardiogram (ECG) signal sensors, pulse wave sensors, and three-axis acceleration sensors to achieve synchronous acquisition and fusion analysis of multi-source physiological signals, significantly improving the accuracy of arrhythmia diagnosis. The main control chip module intelligently processes composite signals, effectively filtering out interference and making diagnostic results more reliable. The Bluetooth module supports real-time data transmission to mobile terminals, facilitating users and doctors to remotely monitor heart health status and achieve all-weather continuous monitoring. The power management module optimizes energy consumption to ensure the long-term and effective operation of the system, taking into account both portability and comfort, and providing an efficient and convenient solution for heart disease prevention and daily management.

[0006] Optionally, the main control chip module includes: an ECG signal filtering module, the ECG signal is used as an input signal, and the ECG signal filtering module is used to pre-process the input signal; an arrhythmia diagnosis module, the arrhythmia diagnosis module is used to diagnose arrhythmia on the subject based on the input signal to obtain the arrhythmia diagnosis result. The present invention pre-processes the original ECG signal through the ECG signal filtering module, effectively filters out noise such as motion artifacts and myoelectric interference, significantly improves the signal-to-noise ratio, and lays the foundation for accurate diagnosis. The arrhythmia diagnosis module uses an algorithm model to perform intelligent analysis of the ECG signal, automatically extracts features and identifies abnormal heart rhythms through a deep learning model, and greatly improves the diagnostic accuracy; it not only ensures real-time processing efficiency, but also improves diagnostic reliability, so that the system can still stably output professional-level diagnostic results in complex environments, providing efficient and accurate technical support for heart health monitoring.

[0007] Optionally, the ECG signal filtering module includes: a filtering algorithm unit, the filtering algorithm unit is used to construct an optimized filtering algorithm based on an adaptive hybrid step size factor and a noise power estimation mechanism; an algorithm experiment unit, the algorithm experiment unit is used to filter the input signal based on the optimized filtering algorithm. The present invention introduces an adaptive hybrid step size factor and a noise power estimation mechanism through the filtering algorithm unit, so that the filtering algorithm can dynamically match signal characteristics and show stronger adaptability and robustness in complex noise environments; the algorithm experiment unit optimizes the filtering algorithm by repeatedly verifying and optimizing it, effectively filtering out noise such as motion artifacts and electromyographic interference, and significantly improving the ECG signal-to-noise ratio; the two units work together to ensure the real-time nature of the filtering process and greatly improve the signal quality, providing a high-fidelity data foundation for subsequent arrhythmia diagnosis, so that the system can ensure the stability of filtering in complex scenarios and significantly enhance diagnostic reliability.

[0008] Optionally, the optimized filtering algorithm is constructed based on the adaptive hybrid step size factor and the noise power estimation mechanism, including: based on the adaptive hybrid step size factor, constructing a dynamic step size factor according to the error signal and the input signal; dynamically adjusting the dynamic step size factor according to the noise power estimation mechanism, and combining the hybrid strategy to construct the optimized filtering algorithm. The present invention constructs an optimized filtering algorithm through a dynamic step size factor, adjusts parameters in real time according to the error signal and the input signal, significantly enhancing the adaptability of the algorithm to signal changes, and the noise power estimation mechanism dynamically optimizes the step size factor, finely adjusts the filtering strength according to the real-time noise level, so that the algorithm maintains efficient denoising capabilities in complex noise environments. Combined with the hybrid strategy, it not only ensures the fidelity of the filtered signal, but also improves processing efficiency, making the filtering process both flexible and accurate, effectively coping with diverse interferences, providing high-quality electrocardiogram signals for subsequent diagnosis, and greatly improving the accuracy of arrhythmia detection and system stability.

[0009] Optionally, constructing a dynamic step factor based on the adaptive hybrid step factor according to the error signal and the input signal includes:

[0010]

[0011] in, is the dynamic step size factor, is the initial step size factor, is the energy of the error signal, is the energy of the input signal, is the noise power estimate, The present invention uses a dynamic step size factor to enable the filtering algorithm to respond to changes in signal and noise in real time. When the noise level is high, the algorithm automatically reduces the step size to avoid overshoot, and when the noise level is low, it increases the step size to accelerate convergence. This significantly improves the stability and accuracy of the filter. It also effectively balances noise suppression and signal fidelity, enabling the algorithm to exhibit stronger adaptability in complex physiological signal environments.

[0012] Optionally, filtering the input signal according to the optimized filtering algorithm includes: obtaining the ECG signal, the pulse wave signal, and the acceleration signal of the subject in a jogging state, the pulse wave signal and the acceleration signal serving as reference input signals; passing the reference input signal through a phase interference cancellation system, and establishing a weight coefficient iterative expression based on an adaptive algorithm to adjust the weight coefficient of the reference input signal; and using the optimized filtering algorithm to iterate the input signal based on the weight coefficient iterative expression to filter the input signal. The present invention collects multi-source physiological signals in a jogging state, uses the pulse wave and acceleration signal as reference inputs, and utilizes the phase interference cancellation system combined with the adaptive algorithm to dynamically adjust the signal weight coefficient to effectively eliminate motion interference; the optimized filtering algorithm iterates based on the weight coefficient to further purify the ECG signal and significantly improve the signal-to-noise ratio. This not only preserves the key features of the ECG signal, but also stably outputs high-quality signals in complex motion scenarios, providing a reliable basis for subsequent arrhythmia diagnosis and making the monitoring system more robust and practical in practical applications.

[0013] Optionally, the step of passing the reference input signal through a phase interference cancellation system and establishing an iterative expression of weight coefficients based on an adaptive algorithm includes:

[0014]

[0015] in, For the The weight coefficient at the iteration, For the The weight coefficient at the iteration, is the dynamic step size factor, For the The input signal at the iteration, For the The error signal at the iteration, is the energy of the input signal, The present invention dynamically adjusts the weight coefficients, enabling the filtering algorithm to respond to changes in the signal and error in real time. The algorithm automatically adjusts the weights to enhance filtering when the error increases, and maintains the weights to preserve details when the signal is stable. This significantly improves the accuracy and stability of adaptive filtering, effectively balancing noise suppression and signal fidelity. This allows the algorithm to exhibit stronger anti-interference capabilities in complex motion scenarios, thereby improving the overall performance of the system.

[0016] Optionally, the arrhythmia diagnosis module includes: a model construction unit, which constructs an arrhythmia classification model based on dynamic convolution and multi-scale feature fusion strategies; a model application unit, which is used to train the arrhythmia classification model according to the overall sample data and obtain an arrhythmia classification result. The model construction unit of the present invention adopts dynamic convolution and multi-scale feature fusion strategies to enable the classification model to adaptively adjust the convolution kernel parameters, accurately capture the transient features in the electrocardiogram signal, and at the same time fuse physiological information of different scales, significantly improving the accuracy of arrhythmia classification; the model application unit is trained through overall sample data to ensure that the model covers a wide range of physiological scenarios and enhances generalization capabilities. The diagnostic system can not only identify complex arrhythmia patterns, but also adapt to individual differences, maintain high performance in diverse clinical data, and provide efficient and accurate intelligent assistance for heart disease screening.

[0017] Optionally, the arrhythmia classification model constructed based on the dynamic convolution and multi-scale feature fusion strategy includes: replacing the one-dimensional convolution layer with a dynamic one-dimensional convolution layer, dynamically adjusting the convolution kernel according to the input signal in combination with the dynamic one-dimensional convolution layer; introducing a multi-scale convolution layer into the convolutional neural network, extracting multi-scale features based on the convolution kernels of different sizes according to the multi-scale convolution layer; fusing the multi-scale features based on the multi-scale feature fusion strategy to construct a fused feature; and using the fused feature as the feature input of the converter module to construct the arrhythmia classification model. The present invention adopts a dynamic one-dimensional convolution layer to adjust the convolution kernel parameters in real time according to the input signal characteristics, significantly enhancing the model's adaptability to signal changes. The design of the multi-scale convolution layer enables the model to simultaneously capture local details and global features, extracting multi-scale information through convolution kernels of different sizes, and improving the richness of feature expression. The feature fusion strategy further integrates multi-scale information to form a more representative fused feature, enabling the model to exhibit higher classification accuracy in complex arrhythmia pattern recognition.

[0018] Optionally, the arrhythmia classification model is trained according to the overall sample data, and the arrhythmia classification result is obtained, including: dividing the overall sample data to construct a training set and a test set, and training the arrhythmia model; obtaining the arrhythmia classification result according to the trained arrhythmia classification model as the arrhythmia diagnosis result. The present invention effectively avoids the risk of overfitting by dividing the overall sample data into a training set and a test set, ensuring that the model has good generalization ability in diversified data, fully optimizes the arrhythmia classification model using the training set, makes the model parameters more consistent with the actual data distribution, significantly improves the classification accuracy, objectively evaluates the model performance based on independent verification of the test set, ensures the reliability of the diagnosis result, improves the model's ability to identify different types of arrhythmias, enhances the practicality of the system in clinical applications, and provides efficient and accurate technical support for cardiac rhythm diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a hardware framework diagram of an arrhythmia diagnosis system based on multi-sensor fusion according to an embodiment of the present invention;

[0020] Figure 2 This is a block diagram of an algorithm implementation of the improved NLMS algorithm according to an embodiment of the present invention;

[0021] Figure 3 This is a signal waveform diagram of a subject in a jogging state according to an embodiment of the present invention;

[0022] Figure 4 A comparison diagram of mean square errors of the filtering algorithms according to an embodiment of the present invention;

[0023] Figure 5 A comparison diagram of filtering waveforms of the filtering algorithm according to an embodiment of the present invention;

[0024] Figure 6 4 is a model structure diagram of an arrhythmia classification model according to an embodiment of the present invention;

[0025] Figure 7 A curve diagram showing changes in the loss function of the arrhythmia model during training according to an embodiment of the present invention;

[0026] Figure 8 This is a curve diagram of the accuracy change of the arrhythmia model during the training process according to an embodiment of the present invention;

[0027] Figure 9 is a confusion matrix diagram of the arrhythmia classification model according to an embodiment of the present invention;

[0028] Figure 10 This is a diagram showing information displayed on a mobile terminal according to an embodiment of the present invention; DETAILED DESCRIPTION

[0029] Specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that these specific details are not necessarily required to practice the present invention. In other instances, well-known circuits, software, or methods are not specifically described to avoid obscuring the present invention.

[0030] Throughout this specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present invention. Therefore, appearances of the phrases "in one embodiment," "in an embodiment," "an example," or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics may be combined in any suitable combinations and / or subcombinations in one or more embodiments or examples. Furthermore, those of ordinary skill in the art will appreciate that the figures provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0031] See Figure 1 According to the design requirements of the portable system, a set of high-performance hardware circuit systems is constructed as a carrier for signal processing by combining signal acquisition, algorithm and wireless transmission. One embodiment of the present invention provides an arrhythmia diagnosis system based on multi-sensor fusion, the system comprising: an electrocardiogram signal module, the electrocardiogram signal module comprising an electrocardiogram signal sensor for acquiring the electrocardiogram signal of a subject; a pulse wave signal module, the pulse wave signal module comprising a pulse wave sensor for acquiring the pulse wave signal of the subject; a three-axis acceleration module, the three-axis acceleration module comprising an acceleration sensor for acquiring the acceleration signal of the subject; a main control chip module, the main control chip module serving as a data processing module of the system for data reception and data analysis, thereby obtaining an arrhythmia diagnosis result; a Bluetooth module, the Bluetooth module serving as a communication module of the system for transmitting the electrocardiogram signal, the pulse wave signal, the acceleration signal and the arrhythmia diagnosis result to a mobile terminal; and a power management module, the power management module comprising charging, power supply and switching circuits for providing power supply for the system.

[0032] In this embodiment, the ECG signal module uses the high-precision ADS1292R chip from Texas Instruments and is connected to the main control chip module via the serial peripheral interface (SPI); the pulse wave signal module is connected to the main control chip module via the integrated circuit bus (I2C); the three-axis acceleration module uses the ADXL345 chip and is connected to the main control chip module via I2C; the main control chip module uses the ultra-low-power STM32L452 series chip from STMicroelectronics; the Bluetooth module uses the V4.0BLE CC2540F256 chip and is connected to the main control chip module via a universal asynchronous receiver / transmitter (UART); the power management module includes charging, power supply and switching circuits, among which the charging circuit uses the BQ21040, and the power supply voltage regulator circuit uses the TLV70033DCK voltage regulator chip. The switching circuit uses a PMOS tube to realize the switching between the system charging and power supply states, thereby reducing the system power consumption and increasing the battery life of the ECG detection device.

[0033] In the arrhythmia diagnosis system based on multi-sensor fusion provided by the present invention, the main control chip module includes:

[0034] A1. ECG signal filtering module. The ECG signal is used as an input signal. The ECG signal filtering module is used to pre-process the input signal.

[0035] The electrocardiogram signal filtering module includes:

[0036] A11, a filtering algorithm unit, wherein the filtering algorithm unit is used to construct an optimized filtering algorithm according to an adaptive hybrid step size factor and a noise power estimation mechanism.

[0037] Research on ECG signal filtering algorithms has achieved some success. However, wearable ECG monitoring devices often experience baseline drift and motion artifacts due to human activity during use. In particular, the frequency of the ECG signal during exercise is similar to the interference frequency, making it difficult to separate. Traditional adaptive filtering methods, such as the LMS and NLMS algorithms, are widely used due to their computational simplicity and good performance. However, they perform poorly when dealing with non-stationary noise and rapidly changing signals. The traditional NLMS algorithm uses a fixed step size factor, which limits the filter's convergence speed and stability in different noise environments.

[0038] In this embodiment, an improved NLMS algorithm is proposed, which introduces an adaptive hybrid step factor and a noise power estimation mechanism. By combining the dynamic step factor, noise power estimation and hybrid strategy, the robustness and adaptability of the algorithm in complex noise environments are improved.

[0039] Specifically, the adaptive hybrid step size factor is dynamically adjusted based on the historical information of the error signal and the energy of the input signal. The current step size is determined by calculating the square of the current error and the square of the historical error in real time, allowing the algorithm to quickly respond to changes in the noise environment.

[0040] Furthermore, to improve the robustness of the algorithm, a noise power estimation mechanism is introduced. The noise power estimation mechanism reduces the step size when the noise is high to avoid over-adjustment; and increases the step size when the noise is low to speed up the convergence, thereby ensuring the stability and efficiency of the filter.

[0041] The dynamic step factor is calculated based on the energy of the current error signal and the energy of the input signal, satisfying the following relationship:

[0042]

[0043] in, is the dynamic step size factor, is the initial step size factor, is the energy of the error signal, is the energy of the input signal, is the noise power estimate, A small positive number.

[0044] The energy of the input signal satisfies the following relationship:

[0045]

[0046] in, is the energy of the input signal, For the The input signal at the iteration.

[0047] It should be noted that A small positive number used to avoid division by zero.

[0048] In this embodiment, a hybrid strategy is proposed by combining the advantages of the dynamic step size factor and the noise power estimation. In each iteration, the direction and amplitude of the weight update are adaptively adjusted according to the energy of the current error signal and the input signal. The dynamic step size factor is dynamically adjusted according to the noise power estimation mechanism. The hybrid strategy is combined to construct an optimized filtering algorithm. The NLMS algorithm is improved to use the adaptive step size factor without significantly increasing the amount of computation.

[0049] See Figure 2 ,The figure shows the algorithm implementation block diagram of the improved NLMS algorithm, ,the pulse wave signal and acceleration signal are used as reference input signals, ,and the ECG signal is used as the original input signal.

[0050] Specifically, the filter output satisfies the following relationship:

[0051]

[0052]

[0053] in, is the filter output, is the sub-filter output, The pulse wave signal is The weight coefficient at the iteration, is the input signal, The acceleration signal is The weight coefficient at the iteration, is the error signal, is the expected output signal.

[0054] A12. Algorithm experiment unit, the algorithm experiment unit is used to filter the input signal according to the optimized filtering algorithm.

[0055] In this embodiment, in order to verify the effectiveness of the improved NLMS algorithm, tests and comparative analysis were conducted through experiments.

[0056] The algorithm experiment unit performs the following steps, including:

[0057] S1. Acquire the electrocardiogram signal, the pulse wave signal, and the acceleration signal of the subject in a jogging state, and use the pulse wave signal and the acceleration signal as reference input signals.

[0058] See Figure 3 ,The figure shows the signal waveform of the subject in the jogging state, including Figure 3 (a) is the ECG signal waveform, Figure 3 (b) is the pulse wave signal waveform and Figure 3 (c) Acceleration signal waveform; collected using an ECG signal sensor, a pulse wave sensor, and an acceleration sensor, with a sampling frequency of 500 Hz.

[0059] S2. Pass the reference input signal through a phase interference cancellation system, and establish an iterative expression of weight coefficients based on an adaptive algorithm to adjust the weight coefficients of the reference input signal.

[0060] Specifically, based on the phase interference cancellation system and combined with the adaptive algorithm, the iterative expression of the weight coefficient is constructed according to the dynamic step size factor to satisfy the following relationship:

[0061]

[0062] in, For the The weight coefficient at the iteration, For the The weight coefficient at the iteration, is the dynamic step size factor, For the The input signal at the iteration, For the The error signal at the iteration, is the energy of the input signal, A small positive number.

[0063] Based on the improved NLMS algorithm and combined with the iterative expression of the weight coefficient, the weight coefficients of the pulse wave signal and the acceleration signal are adjusted to meet the following relationship:

[0064]

[0065]

[0066] in, The pulse wave signal is The weight coefficient at the iteration, The pulse wave signal is The weight coefficient at the iteration, For the The dynamic step size factor at the iteration, is the error signal, is the input signal, represents transpose, is a small positive number, The acceleration signal is The weight coefficient at the iteration, The acceleration signal is The weight coefficient at the iteration.

[0067] In an optional embodiment, the computational complexity of the weight coefficient update process of the LMS algorithm, the NLMS algorithm, and the improved NLMS algorithm is compared to obtain the filtering effect of each algorithm, as shown in Table 1:

[0068] Table 1

[0069]

[0070] For further information, see Figure 4 ,The figure shows the mean square error comparison of the filtering algorithms, including the mean square error learning curves of the LMS algorithm, NLMS algorithm and improved NLMS algorithm.

[0071] Combined with Table 1 and Figure 4A comprehensive analysis was conducted, and the experimental results showed that compared with the NLMS algorithm, the improved NLMS algorithm exhibited higher robustness and adaptability in noise removal. The signal-to-noise ratio (SNR) of the improved NLMS algorithm reached 18.853dB, which was 16% higher than that of the NLMS algorithm and 56.6% higher than that of the LMS algorithm. The improved NLMS algorithm disclosed in the present invention is significantly superior to the LMS and NLMS algorithms. In non-stationary signals and rapidly changing noise environments, the improved algorithm can converge to a steady state more quickly and has a good denoising effect.

[0072] S3. Utilize the optimized filtering algorithm to perform cyclic iteration on the input signal based on the weight coefficient iterative expression to achieve filtering of the input signal.

[0073] In this embodiment, Python is used to design LMS algorithm, NLMS algorithm and improved NLMS algorithm respectively. In the jogging state, the ECG signal containing motion artifact interference and baseline drift noise is iterated using the weight coefficient iterative expression, and the signal is filtered by LMS algorithm, NLMS algorithm and improved NLMS algorithm respectively to obtain the filtered waveform.

[0074] See Figure 5 , the figure shows the comparison of filtering waveforms of filtering algorithms, including the filtering waveforms of LMS algorithm, NLMS algorithm and improved NLMS algorithm.

[0075] Furthermore, a comparative analysis of the weight coefficient updates of the LMS algorithm, NLMS algorithm, and improved NLMS algorithm is performed, as shown in Table 2:

[0076] Table 2

[0077]

[0078] Combine Figure 5 A comprehensive analysis with Table 2 shows that the LMS algorithm has the smallest amount of computation, but is only suitable for environments with limited computing resources. Compared with the LMS algorithm, the NLMS algorithm adds additional multiplication and addition operations, which improves the stability and convergence speed of the algorithm. The improved NLMS algorithm further adds a small amount of multiplication and addition operations. Its dynamic adjustment of the step size factor significantly improves the performance of the algorithm, especially enhances its adaptability and stability in different noise environments.

[0079] A2. An arrhythmia diagnosis module, configured to perform arrhythmia diagnosis on the subject based on the input signal to obtain the arrhythmia diagnosis result.

[0080] The arrhythmia diagnosis module includes:

[0081] A21, a model building unit, wherein the model building unit builds an arrhythmia classification model based on dynamic convolution and multi-scale feature fusion strategy;

[0082] Traditional ECG analysis relies on manual annotation by experts, which is time-consuming, labor-intensive, and susceptible to subjective factors. In recent years, deep learning-based methods have been developed to automatically extract ECG features, enabling efficient and accurate arrhythmia classification. However, the increased model complexity also poses computational resource challenges.

[0083] In this embodiment, dynamic convolution and multi-scale feature fusion strategies are introduced to construct an arrhythmia classification model to reduce the complexity of the model. The device computer uses a 4070ti graphics card and an i7-13700kf processor.

[0084] See Figure 6 , the figure shows the model structure diagram of the arrhythmia classification model; an improved CNN-Transforme is obtained based on CNN-Transforme as the arrhythmia classification model; first, the input layer is used to input the input signal, and the input signal passes through two one-dimensional convolution layers and is batch normalized; secondly, the one-dimensional convolution layer is replaced by a dynamic one-dimensional convolution layer, and the convolution kernel is dynamically adjusted in combination with the dynamic one-dimensional convolution layer to reduce the number of parameters and calculations; then, after the one-dimensional maximum pooling layer, a multi-scale convolution layer is introduced into the convolutional neural network, based on convolution kernels of different sizes, multi-scale features are extracted according to the multi-scale convolution layer to enhance the model representation ability, and based on the multi-scale feature fusion strategy, the multi-scale features are fused to construct fusion features; then, the fusion features are used as the feature input of the converter module to capture the temporal dependency in the signal, and finally, after the flattening layer, the discard layer and two fully connected layers, and output through the output layer.

[0085] A22. A model application unit, which is used to train the arrhythmia classification model according to the overall sample data and obtain an arrhythmia classification result.

[0086] In this embodiment, the arrhythmia classification model is trained by dividing the entire sample data into 80% of the entire sample data as a training set and 20% of the entire sample data as a test set.

[0087] See Figure 7 , the figure shows the loss function change curve of the arrhythmia model during the training process; it shows the change of the loss function of the arrhythmia model as the number of iterations increases, including the training loss curve and the test loss curve.

[0088] See Figure 8, the figure shows the accuracy change curve of the arrhythmia model during the training process; it shows the change of the accuracy of the arrhythmia model as the number of iterations increases, including training accuracy and test accuracy.

[0089] Combine Figure 7 and Figure 8 A comprehensive analysis shows that as the number of iterations increases, the accuracy of the training process continues to rise, but the rate of increase becomes slower and slower, and finally tends to stabilize. The loss function changes in the training process, on the contrary, and finally tends to a stable value. Finally, the accuracy of the arrhythmia model on the training set reaches 99.2%, and the accuracy on the validation set reaches 98%.

[0090] Furthermore, an arrhythmia classification result is obtained based on the trained arrhythmia classification model as an arrhythmia diagnosis result.

[0091] In an optional embodiment, after the training process of the arrhythmia classification model is completed, the performance of the arrhythmia classification model is verified using a test set.

[0092] See Figure 9 The figure shows the confusion matrix diagram of the arrhythmia classification model, which shows the comparison between the predicted labels and the true labels of the five ECG signals N, L, R, A and V by the arrhythmia classification model. The arrhythmia classification results of the arrhythmia classification model are compared with the labels of the MIT-BIH arrhythmia database data, and are expressed in the form of a confusion matrix. The values ​​on the main diagonal represent the number of samples whose true categories are correctly classified as the true categories, and the remaining values ​​are the number of samples that are incorrectly classified.

[0093] Furthermore, the accuracy, precision, recall and F1 value of the arrhythmia classification model for the five types of ECG signals N, L, R, A and V are calculated based on the data in the confusion matrix, and the evaluation index values ​​are obtained, as shown in Table 3:

[0094] Table 3

[0095]

[0096] As shown in Table 3, the classification accuracy reached over 98.6%, the classification precision of each label reached over 95.3%, the recall rate of each label reached over 95%, and the F1 value of each label reached over 96.3%. It can be seen that the improved CNN-Transformer arrhythmia classification model has excellent classification performance, and all indicators are relatively good, basically meeting the requirements for arrhythmia classification prediction.

[0097] Furthermore, in order to verify the advancement and accuracy of the arrhythmia classification model disclosed in the present invention and to better explore the model's improvability, comparative experiments were conducted on the same dataset using a DNN model, a CNN model, and a CNN-Transformer model. The accuracy performance of the three models and the improved CNN-Transformer arrhythmia classification model of the present invention on the same dataset was compared. The accuracy comparison results are shown in Table 4:

[0098] Table 4

[0099]

[0100] As can be seen from Table 4, the improved CNN-Transformer arrhythmia classification model adopted in the present invention has excellent classification performance, indicating that the arrhythmia classification model is more effective in electrocardiogram classification.

[0101] In another optional embodiment, the arrhythmia diagnosis system (multi-sensor diagnostic terminal) based on multi-sensor fusion proposed in the present invention is actually tested and the results are analyzed. In the actual test, experimental data of 10 groups of subjects aged 18-30 were collected, including 7 groups of males and 3 groups of females. All participants had no movement disorders and abnormal heart rates.

[0102] First, the multi-sensor diagnostic terminal was connected to the subject; a pulse wave sensor was worn on the index finger of the left hand; the accelerometer was fixed to the waist because the waist is close to the center of gravity of the human body and can better reflect the overall movement state; and the ECG signal sensor was connected to the electrode patch and worn on the left and right chests and right abdomen respectively; by wearing the pulse wave sensor, ECG signal sensor and accelerometer, 1 minute of ECG signal, pulse wave signal and acceleration signal were synchronously collected and saved as a CSV file; after multi-sensor information fusion and denoising, the data was classified and predicted using the trained improved CNN-Transformer arrhythmia classification model. The ECG signal classification probability results are shown in Table 5:

[0103] Table 5

[0104]

[0105] As can be seen from Table 5, the improved CNN-Transformer arrhythmia classification model proposed in the present invention has better classification prediction performance.

[0106] See Figure 10, the figure shows the information display of the mobile terminal; in order to display vital signs signals in real time, an Android platform APP was developed, which can display ECG signals, pulse wave signals and acceleration signals through the Bluetooth module, and monitor the ECG status in real time, with functions such as signal display, data storage and historical data review.

[0107] In summary, the present invention provides an arrhythmia diagnosis system based on multi-sensor fusion. Against the background of the continuous increase in the incidence of cardiovascular diseases worldwide, a wearable multi-sensor information fusion ECG monitoring system is designed; a hardware acquisition system based on the STM32L452CEU6 chip is designed, and the ADS1292R and ADXL345 chips are used to collect ECG signals, pulse wave signals and acceleration signals, and transmit them to the mobile terminal through the Bluetooth module; an improved NLMS algorithm is proposed, which adopts an adaptive step size factor and noise power estimation mechanism to fuse acceleration and pulse wave signals for filtering, and the signal-to-noise ratio is improved by 16% to 18.853dB; the CNN-Transformer algorithm is improved to construct an improved CNN-Transformer arrhythmia classification model, and dynamic convolution and multi-scale feature fusion strategies are introduced to realize a lightweight model with a classification accuracy of more than 98.6%; test results show that the system is stable, portable, and has low power consumption, which is of great significance for the prevention of cardiovascular diseases and Internet of Things ECG monitoring.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A multi-sensor fusion-based arrhythmia diagnosis system, characterized in that: include: An ECG signal module, comprising an ECG signal sensor for collecting an ECG signal of a subject; A pulse wave signal module, comprising a pulse wave sensor for collecting the subject's pulse wave signal; A three-axis acceleration module, comprising an acceleration sensor for collecting acceleration signals of the subject; A main control chip module, which serves as a data processing module of the system and is used for data reception and data analysis to obtain arrhythmia diagnosis results; A Bluetooth module, serving as a communication module of the system, configured to transmit the electrocardiogram signal, the pulse wave signal, the acceleration signal, and the arrhythmia diagnosis result to a mobile terminal; A power management module, comprising charging, power supply and switching circuits for providing power supply for the system; The main control chip module includes: an ECG signal filtering module, wherein the ECG signal is used as an input signal and the ECG signal filtering module is used to pre-process the input signal; an arrhythmia diagnosis module, configured to perform arrhythmia diagnosis on the subject according to the input signal to obtain the arrhythmia diagnosis result; The electrocardiogram signal filtering module includes: A filtering algorithm unit, configured to construct an optimized filtering algorithm based on an adaptive hybrid step size factor and a noise power estimation mechanism; an algorithm experiment unit, the algorithm experiment unit being used to filter the input signal according to the optimized filtering algorithm; The optimized filtering algorithm is constructed based on the adaptive hybrid step size factor and the noise power estimation mechanism, including: constructing a dynamic step size factor based on the adaptive hybrid step size factor according to the error signal and the input signal; Dynamically adjusting the dynamic step size factor according to the noise power estimation mechanism, and constructing the optimized filtering algorithm in combination with a hybrid strategy; The step of constructing a dynamic step factor based on the adaptive hybrid step factor according to the error signal and the input signal includes: ； in, is the dynamic step size factor, is the initial step size factor, is the energy of the error signal, is the energy of the input signal, is the noise power estimate, A small positive number.

2. The arrhythmia diagnosis system based on multi-sensor fusion according to claim 1, characterized in that: The filtering of the input signal according to the optimized filtering algorithm includes: acquiring the electrocardiogram signal, the pulse wave signal, and the acceleration signal of the subject in a jogging state, and using the pulse wave signal and the acceleration signal as reference input signals; Passing the reference input signal through a phase interference cancellation system, and establishing a weight coefficient iterative expression based on an adaptive algorithm to adjust the weight coefficient of the reference input signal; The optimized filtering algorithm is utilized to perform cyclic iteration on the input signal based on the weight coefficient iterative expression to achieve filtering of the input signal.

3. The arrhythmia diagnosis system based on multi-sensor fusion according to claim 2, characterized in that: The step of passing the reference input signal through a phase interference cancellation system and establishing an iterative expression of weight coefficients based on an adaptive algorithm includes: ； in, For the The weight coefficient at the iteration, For the The weight coefficient at the iteration, is the dynamic step size factor, For the The input signal at the iteration, For the The error signal at the iteration, is the energy of the input signal, A small positive number.

4. The arrhythmia diagnosis system based on multi-sensor fusion according to claim 1, characterized in that: The arrhythmia diagnosis module includes: A model building unit, wherein the model building unit builds an arrhythmia classification model based on a dynamic convolution and multi-scale feature fusion strategy; A model application unit is used to train the arrhythmia classification model according to the overall sample data and obtain an arrhythmia classification result.

5. The arrhythmia diagnosis system based on multi-sensor fusion according to claim 4, characterized in that: The arrhythmia classification model is constructed based on the dynamic convolution and multi-scale feature fusion strategy, including: Replacing the one-dimensional convolution layer with a dynamic one-dimensional convolution layer, and dynamically adjusting the convolution kernel according to the input signal in combination with the dynamic one-dimensional convolution layer; Introducing a multi-scale convolutional layer into a convolutional neural network, and extracting multi-scale features based on the convolution kernels of different sizes according to the multi-scale convolutional layer; Performing feature fusion on the multi-scale features based on the multi-scale feature fusion strategy to construct a fusion feature; The fused features are used as feature inputs to a converter module to construct the arrhythmia classification model.

6. The arrhythmia diagnosis system based on multi-sensor fusion according to claim 4, characterized in that: The arrhythmia classification model is trained according to the overall sample data to obtain an arrhythmia classification result, including: Dividing the overall sample data into a training set and a test set to train the arrhythmia classification model; The arrhythmia classification result is obtained according to the trained arrhythmia classification model as the arrhythmia diagnosis result.

Images