BCG Data Clustering Using New Method So-Called Time-Frequency Moments Singular Value Decomposition (TFM-SVD) and Artificial Neural Networks

Researches during several years showed us that artificial neural networks (ANNs) have strong ability in biomedical field as well as diagnostic applications. They are capable to learn the features of exemplar sets, which is very important whilst under-test process is unknown naturally or there are some difficulties across characterization. For fast or optimally training of ANNs, extracting the most important features of input data is extremely important. In this paper, as innovation we introduced a new method of feature extraction which we called it ´time-frequency moments singular value decomposition (TFM-SVD)´. Then, we used this new method and ANNs for ballistocardiogram (BCG) data clustering to diagnose probable heart disease of six tested subjects. This kind of bio-data has an interesting biomedical recording method via wireless measuring electrodes located on chair surface, sitting area. This feature causes that this measurement and diagnosing method are suitable for using any place such as home, office, etc. The result shows that our introduced method has high performance and not high sensitivity to BCG waveform latency or non-linear disturbance