DEVELOPMENT OF AN IMPROVED LOGISTIC MAPPINGFUNCTION FOR OBJECTIVE ASSESSMENT OF QUALITY OF RECEIVED SPEECH OVER MOBILE TELEPHONE NETWORKS

Abstract

Users’ perspectives approach provides an unbiased assessment of Quality of Service (QoS) of voice offerings in telecommunication networks. This approach is implemented using subjective or objective technique. Though the subjective rating scale is the basis for all ratings, objective technique being computational, can effectively predict quality of degraded speeches. However, existing objective techniques’ mapping functions are unable to properly scale speech quality rating. This study was designed to develop a new logistic mapping function that can effectively predict quality of degraded speeches and provide improved quality rating scale. A speech database consisting of 64 original speeches was developed and transmitted over three mobile telephone networks (A, B and C). Psychoacoustic study was carried out using Zwicker loudness model to evaluate the maximum instantaneous loudness (Nmax) and maximum instantaneous loudness level (Lmax) of original and received speeches. The quality of received speeches relative to the transmitted speech was obtained using Nmax and Lmax values. Subjective quality test was carried out on the received speeches based on listening-only technique and Absolute Category Rating (ACR). Objective quality assessment was carried out using Perceptual Evaluation of Speech Quality (PESQ) model, and the scores obtained were mapped using the ITU-T P.862.1 mapping function. The mapped and subjective scores were compared to obtain the correlation coefficient (r), the prediction error (EP), and the Root Mean Square Error (RMSE). A new logistic mapping function for PESQ was developed by optimising the steepness of the logistic S-curve to obtain the growth rate that maximised the range of the quality score. The new mapping function was compared with two international standard mapping functions (ITU-T P.862.1 and Morfitt III–Cotanis). Data were analysed using ANOVA at α_0.05. The Nmax of 46.19 sone and Lmax of 95.29 phon were obtained for the transmitted speech, and Nmax of 19.65, 17.13, 16.46 sones and Lmax of 82.97, 80.98, 80.41 phons were obtained for the received speeches over A, B, and C networks, respectively. The relative quality of the received speeches for the Nmax and Lmax were 42.55, 37.08, 35.64% and 87.06, 84.98, 84.37%, respectively. The subjective test of received speeches over networks A, B, and C, resulted in 2.902±0.380, 2.952±0.447 and 2.983±0.612, respectively, while the objective mapped scores were 2.615±0.563, 2.589±0.594 and 2.693±0.730, respectively. Comparing the mapped and subjective scores produced r of 0.854, 0.871, and 0.848, EP of 0.4264, 0.4724 and 0.4825, and RMSE of 0.4230, 0.4687 and 0.4787, respectively. The optimised steepness resulted in growth rate of 2.2106 and quality coverage of about 1.005 to 4.950. Score coverages of 98.6, 86.8 and 93.7% of the subjective scale were obtained with the developed logistic mapping function, ITU-T P.862.1 and Morfitt III –Cotanis mapping functions, respectively. This indicates a significant improvement over the other two mapping functions. A new logistic mapping function that enhanced objective technique for users’ perspective approach in the assessment of speech quality over mobile telephone networks was developed.