Human activity recognition (HAR) is becoming increasingly important, especially with the growing number of elderly people living at home. However, most sensors, such as cameras, do not perform well in low-light environments. To address this issue, we designed a HAR system that combines a camera and a millimeter wave radar, taking advantage of each sensor and a fusion algorithm to distinguish between confusing human activities and to improve accuracy in low-light settings. To extract the spatial and temporal features contained in the multisensor fusion data, we designed an improved CNN-LSTM model. In addition, three data fusion algorithms were studied and investigated. Compared to camera data in low-light environments, the fusion data significantly improved the HAR accuracy by at least 26.68%, 19.87%, and 21.92% under the data level fusion algorithm, feature level fusion algorithm, and decision level fusion algorithm, respectively. Moreover, the data level fusion algorithm also resulted in a reduction of the best misclassification rate to 2%~6%. These findings suggest that the proposed system has the potential to enhance the accuracy of HAR in low-light environments and to decrease human activity misclassification rates.

Arrhythmias are among the diseases with high mortality rates worldwide, causing millions of deaths each year. This underscores the importance of real-time electrocardiogram (ECG) monitoring for timely heart disease diagnosis and intervention. Deep learning models, trained on ECG signals across twelve or more leads, are the predominant approach for automated arrhythmia detection in the AI-assisted medical field. While these multi-lead ECG-based models perform well in automatic arrhythmia detection, their complexity often restricts their use on resource-constrained devices. In this paper, we propose an efficient, lightweight arrhythmia classification model using a knowledge distillation technique to train a student model from a teacher model, tailored for embedded intelligence in wearable devices. The results show that the student model achieves 96.32% accuracy, which is comparable to the teacher model, with a remarkable compression ratio that is 1242.58 times smaller, outperforming other lightweight models. Enabled by the proposed model, we developed a wearable ECG monitoring system based on the STM32F429 Discovery kit and ADS1292R chip, achieving real-time arrhythmia detection on small wearable devices.

Noise is a common problem in wearable electrocardiogram (ECG) monitoring systems because the presence of noise can corrupt the ECG waveform causing inaccurate signal interpretation. By comparison with electromagnetic interference and its minimization, the reduction of motion artifact is more difficult and challenging because its time-frequency characteristics are unpredictable. Based on the characteristics of motion artifacts, this work uses adaptive filtering, a specially designed ECG device, and an Impedance Pneumography (IP) data acquisition system to combat motion artifacts. The newly designed ECG-IP acquisition system maximizes signal correlation by measuring both ECG and IP signals simultaneously using the same pair of electrodes. Signal comparison investigations between ECG and IP signals under five different body motions were carried out, and the Pearson Correlation Coefficient |r| was higher than 0.6 in all cases, indicating a good correlation. To optimize the performance of adaptive motion artifact reduction, the IP signal was filtered to a 5 Hz low-pass filter and then fed into a Recursive Least Squares (RLS) adaptive filter as a reference input signal. The performance of the proposed motion artifact reduction method was evaluated subjectively and objectively, and the results proved that the method could suppress the motion artifacts and achieve minimal distortion to the denoised ECG signal.

Human activity recognition (HAR) can effectively improve the safety of the elderly at home. However, non-contact millimeter-wave radar data on the activities of the elderly is often challenging to collect, making it difficult to effectively improve the accuracy of neural networks for HAR. We addressed this problem by proposing a method that combines the improved principal component analysis (PCA) and the improved VGG16 model (a pre-trained 16-layer neural network model) to enhance the accuracy of HAR under small-scale datasets. This method used the improved PCA to enhance features of the extracted components and reduce the dimensionality of the data. The VGG16 model was improved by deleting the complex Fully-Connected layers and adding a Dropout layer between them to prevent the loss of useful information. The experimental results show that the accuracy of our proposed method on HAR is 96.34%, which is 4.27% higher after improvement, and the training time of each round is 10.88 s, which is 12.8% shorter than before.