Objective and Subjective Evaluations of Adaptive Noise Cancellation Systems with Selectable Algorithms for Speech Intelligibility

Abstract

Adaptive Noise Cancellation (ANC) systems with selectable algorithms refer to ANC systems that are able to change the adaptation algorithm based on the eigenvalue spread of the noise. These systems can have dual inputs based on the conventional ANC structure or a single input based on the Adaptive Line Enhancer (ALE) structure. This paper presents a comparison of the performance of these two systems using objective and subjective measurements for speech intelligibility. The parameters used to objectively compare the systems are the Mean Square Error (MSE) and the output Signal to Noise Ratio (SNR). For subjective evaluation, listening tests were evaluated using the Mean Opinion Score (MOS) technique. The outcomes demonstrate that for both objective and subjection evaluations, the single input ALE with selectable algorithms (S-ALE) system outperforms that of the dual input ANC with selectable algorithm (S-ANC) in terms of better steady-state MSE by 10%, higher SNR values for most types of noise, higher scores in most of the questions in the MOS questionnaire and a higher acceptance rate for speech quality.