Анализа обележја у говорном сигналу за потребе препознавања мултимодалног говора

Abstract

У системима аутоматског препознавња говора све значајнију улогу игра и мултимодални говор...

In automatic speech recognition systems, multimodal speech is playing an increasingly important role. Different modes of speech involve different acoustic features. Particular importance is given to normal (natural) speech, whisper and their combinations (when the system is trained by normal speech and then used to recognize whisper, and vice versa). The match scenarios: „normal/normal“ and „whisper/whisper“, and mismatch scenarios: „normal/whisper“ and „whisper/normal“ are considered here...