EEG based hearing threshold classification using fractal feature and neural network

In this paper, proposed a method to classify EEG time series signals recorded from left and right ears by presenting an acoustical stimulus in a time locked manner. Fractal dimensional features were extracted in order to measure the complexity of temporal dynamics and response onset of auditory evoked potentials of normal hearing and abnormal hearing persons. This study identified a significant potential difference between fractal dimensional values of the normal hearing and abnormal hearing person. The extracted fractal features were then associated to the hearing threshold perception and a neural network model for left and right ears were developed. The classification results in discriminating the left and right ear of normal and abnormal person was reported as 90% and 95% with specificity of 90%, sensitivity of 100%. Since the results were promising, it can be safely adopted in screening the hearing threshold level of a person in clinics.