Hand Tremor Classification Using Bispectrum Analysis of Acceleration Signals and Back-Propagation Neural Network.

This paper presents a new approach to classify three types of tremor, including parkinsonian, essential and physiological tremors, by using bispectrum analysis of time series of hand tremor and neural network. The acceleration signals of hand tremor from voluntary subjects were recorded and the features of diagonal slices of bispectrum were extracted. A simple BP artificial neural network classifier based on LM algorithm has been used for classification. The study indicates the accuracy rate is over 92.9%. The results show that the method has a better performance than other methods, such as time or frequency domain analysis,and provide a new approach to classify tremor for clinical neurosurgeon. [ABSTRACT FROM AUTHOR]