Aiming at the demand for precise regulation of exercise intensity due to individual differences in sports health management, this paper proposes a data acquisition device for sports health signs based on the MICNN model. The device collects data through a combination of multiple sensor modules and relies on data filtering for data processing. The MICNN model is used to construct a multi-sensor feature parallel extraction architecture, and the exercise heart rate prediction is realized through feature aggregation. The designed device is put into experiments, and a Butterworth low-pass filter is used to process the skin electrical signals and extract the timedomain features containing SCR and SCL. The heartbeat signals are normalized and the validity of respiratory signal acquisition is verified using polysomnography. Exercise heart rate was predicted by the MICNN model, and its performance was evaluated by a combination of Bland-Altman analysis and comparative experiments. The mean value of model-predicted heart rate deviation was 0.03, the 95% agreement range (±1.96 times standard deviation) was +4.52bpm and -4.46bpm, and the RMSE, MAE, and MAPE were 0.73, 0.52, and 7.25%, respectively, which were significantly lower than those of other control models.
