Auditory Non-Intrusive Quality Estimation Plus (Anique+): Perceptual Model for Non-Intrusive Estimation of Narrow-Band Speech Quality (Includes Access to Additional Content)
The objective speech quality estimated by the Auditory Non-Intrusive QUality Estimation Plus (ANIQUE+) model of this American National Standard (ANS) is the subjective quality of telephone band speech. The quality estimated or predicted by the ANIQUE+ model is not conversational quality but listeningonly quality@ which can be originally obtained by subjective auditory tests investigating listening quality in Absolute Category Rating (ACR) scale using a common receiving handset at a standard listening level of 79 dB SPL (See ITU-T P.800 and P.830). This ANS can be used to predict the quality of one-way speech transmission. Purpose The need for objective models for estimating speech quality is becoming more prominent@ especially in complex modern telecommunication network environments. In addition to existing traditional public switched telephone networks (PSTN)@ various types of mobile and internet-based networks are being widely used. The resulting interconnected heterogeneous network increases significantly the number of factors that can affect speech quality@ and the understanding of the impact of individual system components and combinations of them on the end-to-end speech quality is very difficult. In addition@ these newer networks provide a tradeoff between service quality and cost. Thus@ the reliable estimation of speech quality over modern telecommunication networks is very important not only for network systems design and development@ but also for the maintenance of quality of service (QoS). This ANS defines an algorithm which provides acceptable accuracy in estimating the quality of speech processed by telecommunication networks in a non-intrusive manner. For 20 Mean Opinion Scope (MOS) test databases which have never been used in the development of objective models@ the average correlation between subjective and objective quality scores is about 0.87 for this ANS@ whereas the ITUT P.563 shows about 0.81 correlation for the same task.