Alcoholism diagnosis from EEG signals using continuous wavelet transform

Effective working of brain computer interface largely depends upon mental state and vigilance level of human brain. Electroencephalogram signal undergoes for unpredictable changes when vigilance state of human brain alters widely and sometimes cause wrong interpretation by brain computer interface during operation. Hence, brain computer interface needs to investigate subject´s brain alertness level frequently to avoid false command generation. In present work, an approach of feature extraction from electroencephalogram signals using continuous wavelet transform is proposed and validated for identification of the two different brain states i.e. alcoholism and normal. The coefficients of continuous wavelet transform exploiting four distinct base wavelets are computed from processed electroencephalogram signals. Further, statistical parameters are calculated for each base wavelet and employed to prepare feature vector from electroencephalogram. The prepared feature vector is used to perform training and validation of the soft computing techniques in present work, which includes support vector machine, neural network and random forest tree classifier. A comparative study is performed for different feature vectors in combination of soft-computing techniques to obtain effective methodology for alcoholism diagnosis.