Using Automatic Speech Recognition for Speech Comprehension Evaluation in the Cochlear Implant

The cochlear implant (CI) is a sophisticated electronic device designed to partially restore hearing for individuals with severe-to-profound hearing loss. To assess speech intelligibility in CI listening, traditional objective evaluation methods obtain scores from psychoacoustic weights across frequency bands but do not account for auditory processing in speech comprehension. In this paper, we propose a novel objective evaluation approach for the CI using Whisper, an automatic speech recognition (ASR) system. Speech comprehension is estimated by assessing the word error rate (WER) and character error rate (CER). The proposed measure is used to evaluate CI simulation speech processed using the common Advanced Combination Encoder (ACE) strategy and the deep-learning based ElectrodeNet-CS strategy. The experimental results demonstrate the feasibility of using ASR technology for evaluating CI speech perception, eliminating the need for original clean speech. This innovative study proposes a unique AI-based alternative for objective evaluation and offers new insights into estimating CI speech comprehension.