Limitations of perceptual evaluation of speech quality on VoIP systems

This paper presents the shortcomings of perceptual evaluation of speech quality, an objective method of end-to-end voice quality measurement for narrow band telephone networks and speech codec. The main focus of the paper is to discuss practical problems associated with perceptual evaluation of speech quality method in evaluating jitter buffer performance and quality assurance testing of VoIP terminals. The majority of the limitations of PESQ were observed during voice quality tuning experiments for VoIP feature. In our implementation, VoIP link is provided in the Wi-Fi mode of Unlicensed Mobile Access system. AMR NB is the speech codec used in VoIP terminal to compress and decompress speech as preferred by 3GPP standards. The experiments were carried out in controlled lab environment which is well suited for speech quality assessment.