Non-intrusive assessment of perceptual speech quality using a self-organising map

A new output-based method for non-intrusive assessment of speech quality for voice communication system is proposed and its performance evaluated. The method is based on comparing the output speech to an appropriate reference representing the closest match from a pre-formulated codebook containing optimally clustered speech parameter vectors extracted from a large number of various undistorted clean speech records. The objective auditory distances between vectors of the distorted speech and their corresponding matching references are then measured and appropriately converted into an equivalent subjective score. The optimal clustering of the reference codebook is achieved by a dynamic k-means method. A self-organising map algorithm is used to match the distorted speech vectors to the references. Speech parameters derived from Bark spectrum analysis, Perceptual Linear Prediction (PLP), and MelFrequency Cepstral coefficients (MFCC) are used to provide speaker independent parametric representation of the speech signals as required by an output-based quality measure.