This paper designs an overall scheme for a wearable multi-sensor physiological monitoring system, which monitors physiological parameters such as electrocardiogram signals and blood pressure. It combines CNN, LSTM, and MHA and optimizes them using the NRBO algorithm to construct a CNN-LSTM-MHA hybrid neural network model. The performance of this model in heart rate monitoring and blood pressure prediction is evaluated through experiments. In terms of heart rate detection and blood pressure prediction, the CNN-LSTM-MHA model demonstrates the best overall performance and exhibits superior robustness.