Speech quality in various acoustic environments have been evaluated using subjective speech quality measures such as the mean opinion score (MOS) and the diagnostic rhyme test (DRT). However, such subjective measures are very expensive and laborious in assessing speech quality. Therefore, it would be desirable to have an objective speech quality measure which can estimate the subjective quality of speech signal in the cellular communication network.
Several objective measures such as the segmental signal to noise ratio (SegSNR), the linear predictive coefficient-cepstral distance (LPC-CD), the bark spectral distance (BSD), and the perceptual speech quality measure (PSQM) have been proposed for estimating the subjective speech quality. However, they do not perform so good as objective measures for speech quality in the CDMA system because the speech signal in the CDMA system becomes distorted by various channel errors which are not considered in those objective measures. Among the channel distortions which result in severe performance degradation in the CDMA system is the frame erasure. In addition to the channel error distortion, the time alignment between the original and the distorted signals is very critical to objective quality measures because many of them are very sensitive to the accuracy of the synchronization.
In this thesis work, we propose an objective measure of speech for the CDMA system. We investigate the effect of the frame erasure and the time synchronization between the original and the distorted speech signals. From the experimental results, we develop a new efficient objective measure FE-PSQM which can take into account the effect of the frame erasure in the CDMA system and a new synchronization algorithm. Exhaustive experiments show that the proposed objective measure and the synchronization algorithm result in improved performance compared with that of conventional methods, especially in the environments of high frame erasure rate.
