Phonetic-acoustic database of trigrams for Russian dialects speech recognition

The paper describes the development of a phonetic-acoustic database of Russian trigrams applicable for dialects speech recognition. Trigrams were defined as sets of phonetic-acoustic elements in the developed database. The phonetic-acoustic database aims to collect various pronunciations including dialects and other human speech defects as it’s the key problem for up-to-date recognition systems. The scales for speech recognition based on trigrams include frequency of their use, positions in words and associative force of each trigram.