Design and development of an intelligent hearing ability level assessment system using somatosensory stimuli

Abstract

Hypoacusis is the most prevalent sensory disability in the world which leads to impeding speech in human beings. One best approach to tackle this issue is to conduct early and effective hearing screening test using Electroencephalogram (EEG). Auditory evoked potential (AEP) is a type of EEG signal emanated from the brain scalp by presenting an acoustical stimulus in a time-locked manner. AEP response reflects the auditory ability level of an individual. In this thesis, an intelligent hearing ability level assessment system is designed to determine the hearing threshold response and hearing perception response using AEP signals. An objective method that records the complete characteristics of the AEP signals to determine the hearing responses at low stimulation intensity (20 dB) is analyzed. Two simple and AEP based hearing protocols are developed to determine the significant correlations between the brain dynamics and the auditory responses. Firstly, the AEP based hearing threshold response protocol has been proposed to detect the hearing sensitivity level of the normal hearing and abnormal hearing subjects. Secondly, the AEP based hearing perception response protocol has been proposed to determine the different hearing perception levels (20 dB, 30 dB, 40 dB, 50 dB and 60 dB) of the normal hearing subjects. Simple preprocessing algorithms are presented to remove noise from the raw signals. New hearing-threshold factors using autoregressive pole-tracking algorithms are applied to extract the lower and upper hearing-threshold factors of a subject. Three spectral features and three fractal features are proposed and tested with classifiers. A particle swarm optimization based algorithm is proposed to train the neural networks. From the results, for the normal hearing subjects, the maximum hearing-threshold lower (HL) values for the left and right ears are observed as 6.995 and 7.439 respectively, and the maximum hearing-threshold upper (HU) values for the left and right ears are observed as 9.501 and 9.997 respectively. For the abnormal hearing subjects, the maximum HL values for the left and right ears are determined as 10.610 and 11.038 respectively, and the maximum HU values for the left and right ears are determined as 15.594 and 15.698 respectively. From the results, it is inferred that for abnormal hearing participants, the hearing threshold values are almost 30-40% higher than the normal hearing participants. Further, higuchi fractal feature (HFF) algorithm using particle swarm algorithm based neural network (PSONN) for the hearing frequency level of 1000 Hz has achieved the overall maximum classification accuracy of 95% and 97.5% for the left and right ears, respectively. Furthermore, it is also inferred that an auditory frequency of 1000 Hz has the predominant acoustic characteristics that can be used as a critical frequency to determine the hearing-threshold of normal and abnormal hearing subjects. In addition, the HFF algorithm using PSONN for the hearing frequency level of 8000 Hz has achieved the maximum classification accuracy of 88.57% and 91.42% for the left and right ears in discriminating the five different hearing perception levels. Furthermore, it can be noticed that the significant increase in the hearing perception level along with the stimulus intensity levels. The results obtained were promising with the experimental data; it can be used to detect the hearing states for newborns, infants, and multiple handicaps, person who lacks verbal communication and behavioral response to the sound stimulation.