Radon Transform of Auditory Neurograms: A Robust Feature Set for Phoneme Classification

What is it about?

In this study, a neural-response-based feature extraction method was proposed for a robust phoneme classification system. Features were extracted by applying a well-known Radon transform on the neurogram. The performance of the system using the proposed neural feature was evaluated in quiet and under noisy conditions and also compared to the classification accuracy from several existing methods.

Featured Image

Why is it important?

Automatic speech recognition (ASR) systems need to be robust to extrinsic variations in the speech signal, such as background noise, reverberation, and transmission-channel noise to be applicable in many real-life scenarios. Thus, designing a robust phoneme classifier is of utmost importance for developing a robust ASR system. The proposed neural-response-based approach outperformed all other baseline features in clean, with additive noise and telephone-channel distortion conditions.