Heart rate prediction based on CNN-Transformer model

With the aging population in our society, there is a growing demand for the detection of vital signs. Millimeter-wave radar has gained significant attention due to its high accuracy and non-contact nature. To enhance the accuracy of heart rate prediction, this paper proposes a method based on a hybrid CNN-Transformer model. In this method, the CNN layer is utilized to extract local features of the heartbeat phase, compensating for the limited information mining capability of the Transformer encoder when dealing with multivariate datasets. Concurrently, the Transformer module is employed to capture dependencies between local and global features in the sequence data, aiding the model in better understanding the overall characteristics of the heartbeat signal. The experiments were conducted using an FM continuous wave radar device with a center frequency of 77 GHz. The experimental results demonstrate that the proposed model achieves an average absolute error of only 0.94 bpm, outperforming existing models when compared to medical-grade ECG acquisition equipment. This validates its superiority in heart rate prediction.