Our laboratory has developed the technology for person's voice identification by the input arbitrary speech audio records. The main idea consists in the sequential comparison of the input speech audio record's parametric description with the individual speech samples being synthesized on the text of the input audio record. For generating information about the phonetic content of the speech signal, the Multi-voiced TTS system is used. The non-linear time comparison of natural and synthesized speech signals is performed using the software specially developed on the basis of dynamic programming algorithms. Such analysis-by-synthesis principle allows to automate real-time search and identification of a person's voice in a very large-scale specialized acoustic databases (VLS SADB).