Evaluation of the Quality of Music Signals Limited by the Frequency Band

In this paper, the results of the musical signals quality assessment, with the use of objective and subjective measures, are presented. The topics covered in this paper are relevant because the growing needs of users of communication lines leading to the gradual increasing demands for bandwidth that is given to the signals. Nowadays, frequency band 300-3400 Hz (Narrow Band) in analog and mobile telephony is used. However, with the development of mobile technologies and network communications, it becomes possible to expand the bandwidth to WideBand (50-7000 Hz), Super-WideBand (50-14000 Hz) and even to Full-Band (20-20000 Hz). Such an extension of the frequency band can be explained by the desire to transmit musical signals on communication channels in a qualitative way. On the other hand, the average person may not notice the difference between two complex musical works, the spectrum of one of which extends to 20 kHz, and for another is limited to 14 kHz or even 7 kHz. Subjective music signals quality assessment was carried out by 23 students with an average age of 22 years, without hearing impairments. For objective evaluation, 4 quality measures were used, including segment signal-to-noise ratio, log-spectral distortions, bark-spectral distortions, as well as "perceptual evaluation of audio quality" measure specifically designed to evaluate the quality of music signals. The last measure is of particular interest, since the algorithm of its calculation takes into account the features of the human auditory system. The validity of the results of previous studies is confirmed, where it was stated that the frequency band of 12-14 kHz is sufficient to ensure that the musical signal is not different from the reference signal by ear. However, it should be borne in mind that the indicated result is true on average, so it is not excluded that individual listeners will assume that such a band is not wide enough for the quality reproduction of musical signals. In view of this, the results obtained in this article should be expanded, in the future, by the law of the distribution of the bandwidth limits. The examples of the "segmental signal-to-noise ratio" and “logarithmic spectral distortion” indicators demonstrated the need to take into account the peculiarities inherent in certain objective quality measures, as well as the need to provide a sufficiently large statistical volume of various musical material to obtain reliable estimates of the quality of musical signals. In particular, when using the measure "segmental signal-to-noise ratio," it is advisable to increase the sampling rate by 2-4 times, using the interpolation of the compared signals. Correspondence maps for subjective and objective measures of quality are constructed, which allows calibrating of objective estimation systems of musical signals quality. In contrast to the correlation coefficient, the use of correspondence maps suggests the existence of a nonlinear relationship between subjective and objective estimates, which contributes to improving the accuracy of the evaluation. Ref. 13, fig. 3, tabl. 2.