?
Method for testing the stability of an autoregressive model of the vocal tract and adjusting its parameters
Within the framework of the traditional scope of investigations in the field of acoustic measurements, we
consider an autoregressive model of the vocal tract, which is a key link in the speech apparatus of human
beings. We mention the existence of an urgent problem of guaranteeing the stability of the autoregressive model
in systems with adaptation of their parameters to the observed speech signals of short duration. To overcome
this difficulty, we pose the problem of testing the stability of the autoregressive model and adjustment of its
parameters according to the results of testing. The required investigations are based on the original authors’
technique of the formant analysis of vowel sounds of speech via the synthesis of a recursive shaping filter
in the mode of free oscillations. For the solution of the posed problem, we propose a procedure aimed at
testing the stability of the autoregressive model of vocal tract and adjustment of its parameters. The method
is based on a two-stage algorithm of transformation of the autoregressive model of vocal tract. In the first
stage of transformation, the stability of the autoregressive model is checked according to the impulse response
of the shaping filter. In the second stage, if the stability of the autoregressive model is violated, its impulse
response is modified as a result of the element-by-element multiplication by a variable exponential quantity
asymptotically convergent to zero.We develop a regular algorithm for recalculating a modified impulse response
into the adjusted vector of autoregressive parameters in the second stage of transformation. According to the
results of experimental verification of the proposed method, we make a conclusion that the guaranteed stability
of the autoregressive model of the vocal tract is attained with minimal distortions in the frequency domain.
The obtained results can be useful for the development and improvement of the systems of automatic speech
recognition, digital speech communications, artificial intelligence, and other information systems based on the
use of data compression and speech encoding according to the autoregressive model of the vocal tract in the
course of automatic processing of speech signals.