Modeling Heart Rate Variability with a HMM-based Neural Network
This paper proposes a method of modeling heart rate variability combining wavelet transform with a neural network based on a hidden Markov model. The proposed method has the following features: 1. The wavelet transform is used for feature extraction to extract the local change of heart rate variability in the timefrequency domain. 2. A new recurrent neural network incorporating a hidden Markov model is used to model the different patterns of heart rate variability caused by individual variations, physical conditions and so on. In experiments, five subjects were subjected to a mental workload, and the proposed method was used map subjective rating scores of their mental stress and the pattern of heart rate variability. Experiments confirmed that the proposed method achieved highly accurate modeling.