Jan Holub, Michael Street, Ondrej Tomiska
COMPUTATIONALLY EFFICIENT NON-INTRUSIVE ALGORITHM FOR SPEECH TRANSMISSION QUALITY MEASURMENT
A new, wavelet-based, non-intrusive method for speech transmission quality measurements is described in the abstract. It models human perception of quality of transmitted speech signal. The deployed Discrete Wavelet Transform, in comparison with Fourier Transform, enables to reduce the computational power. In comparison with standardized methods (based on ITU-T P.563 algorithms), the described method saves about 90% of operations needed, achieving about 90% of the results of P.563. Thus, it is suitable for operational continuous assessment, or even for applications embedded in the mobile terminal.