Evaluation of heart condition based on ballistocardiogram classification using compactly supported wavelet transforms and neural networks

One of the most usual causes of death of the human are among heart diseases. Several electronic devices have been developed to assist clinicians in monitoring and diagnosing heart diseases. Ballistocardiography (BCG) was one of popular methods before the 1970s but after that other methods have replaced it, partly because the devices were difficult to construct. Recently developed sensors offer new unobtrusive possibilities to evaluate the condition of the patient's heart even at home without attaching electrodes to the patient. Thus, it is suitable for evaluation of the heart condition in any place because of being user-friendly method. In this study, we applied compactly supported (Daubechies as well as biorthogonal) wavelet transforms in a comparison way to extract essential features of the BCG signal and neural networks to classify the BCG. Initial tests with BCG from six subjects indicate that the method can classify the subjects to three classes with a high accuracy. The method is almost insensitive to latency and non-linear disturbance. Moreover, the wavelet transform requires no prior knowledge of the statistical distribution of data samples and the computational complexity and training time are reduced