Return Home

A true measure of speech quality is obtainable by subjective listening tests. These involve a panel of trained listeners and are impractical for most applications. Objective listening quality algorithms have been developed and improved over the years. These algorithms can be categorized into parametric and psychoacoustic algorithms. Parametric algorithms have been widely used for estimation of the speech quality in wireless networks based on parameters such as the SNR, BER, and FER. The indirect estimation of the speech quality based on these parameters often leads to unreliable quality which is compensated through excessive transmission power. Psychoacoustic algorithms, on the other hand, estimated the speech quality based on elaborate models human hearing and directly from the speech signal. Although, the results from psychoacoustic algorithms closely match those of subjective quality tests, their use has been limited to off-line applications.

The objectives of this subproject are to investigate application of psychoacoustic algorithms for monitoring and control of speech quality in modern wireless networks as well as the potential benefits in resource management in such networks.