?
Метод преобразования речевого сигнала для улучшения разборчивости речи
The problem of improving speech intelligibility in voice communication systems is considered. The acute issue of speaker recognition when applying known methods for solving this problem is highlighted. To overcome the specified problem, a new method for transforming the speech signal based on an autoregressive model of the vocal tract and the principle of frequency-selective amplification of main formants is proposed. An example of practical implementation of the new method using fast Fourier transform algorithms is presented. Estimates of computational costs and performance are provided. A field experiment was set up and conducted. According to its results, the positive effect achieved by applying the proposed method has been established, namely: increasing the intelligibility of speech from a control speaker while maintaining a sufficiently high degree of recognizability of his voice. The obtained results are intended for use in developing new and modernizing existing voice communication systems, including mobile communications and VoIP systems.