Optimization of a Real-Time Wavelet-Based Algorithm for Improving Speech Intelligibility

TL;DR

This paper presents a real-time wavelet-based algorithm that enhances speech intelligibility by adjusting sub-band gains, improving transcription accuracy under various noise and hearing loss conditions, with applications in hearing aids and speech processing.

Contribution

It introduces a simplified, real-time wavelet-based method for speech enhancement that effectively improves intelligibility across different noise levels and hearing impairments.

Findings

16.9% increase in transcription accuracy for clean speech

9.5% increase in transcription accuracy for noisy speech

Universal sub-band gains effective up to 4.8 dB noise-to-signal ratio

Abstract

The optimization of a wavelet-based algorithm to improve speech intelligibility along with the full data set and results are reported. The discrete-time speech signal is split into frequency sub-bands via a multi-level discrete wavelet transform. Various gains are applied to the sub-band signals before they are recombined to form a modified version of the speech. The sub-band gains are adjusted while keeping the overall signal energy unchanged, and the speech intelligibility under various background interference and simulated hearing loss conditions is enhanced and evaluated objectively and quantitatively using Google Speech-to-Text transcription. A universal set of sub-band gains can work over a range of noise-to-signal ratios up to 4.8 dB. For noise-free speech, overall intelligibility is improved, and the Google transcription accuracy is increased by 16.9 percentage points on average and 86.7 maximum by reallocating the spectral energy toward the mid-frequency sub-bands. For speech already corrupted by noise, improving intelligibility is challenging but still realizable with an increased transcription accuracy of 9.5 percentage points on average and 71.4 maximum. The proposed algorithm is implementable for real-time speech processing and comparatively simpler than previous algorithms. Potential applications include speech enhancement, hearing aids, machine listening, and a better understanding of speech intelligibility.