Active and Passive Oddball Paradigm for Automatic Speech Discrimination Assessment

Abstract

An oddball paradigm is an experimental design that uses a sequence of one repeating stimulus called the standard stimulus. This sequence is infrequently interrupted by a different stimulus called the deviant or target stimulus. Potentially the oddball paradigm can be employed in an EEG-based speech discrimination assessment protocol. Speech discrimination indicates how well a person can differentiate between different words. Analyzing EEG measurements such as the Event-Related Potentials (ERPs) may help to achieve the goal of automated assessment process. In this work we compare two listening modes in an oddball paradigm in order to find a suitable mode for assessing speech discrimination automatically. The two listening modes include passive and active listening. Passive listening is when the listener does not pay attention to what they hear. Active listening is when the listener actively pays attention to the sound. We tested these two listening modes using two Thai words with consonant contrast. We compared the ERP waveform, classification accuracy, and attention during passive and active listening. We found that passive listening produced clearer ERP waveform. However, active listening achieved higher accuracy and engaged less attention. Therefore, we recommend using active listening for an auditory oddball paradigm when assessing speech discrimination.