A combined wideband speech and audio coder using human articulatory and auditory models
Guangyu Wang · The Journal of the Acoustical Society of America · 1999
The modern high-quality speech and audio coding algorithms are mostly based on human articulatory models. On the other hand, high-quality audio coders are mostly based on models of human auditory system. However, neither of these two methods can provide acceptable performance for both music and speech. In this paper a coder is proposed which operates for both wideband speech and audio signal. The structure of the coder is based on both human articulatory and auditory models. In the first part the LPC parameters are drawn using the conventional autocorrelation method in order to simulate the human articulatory system. With LPC parameters the redundancy of input signal in time domain is removed. In the second part of the coder, the LPC residual signal is further transformed into frequency domain through MLT (modulated lapped transform). The MLT coefficients are then quantized in frequency domain using the human auditory properties such as masking property. The informal listening test has shown that the proposed coder provides very good quality for wideband speech and for most music signals.