Speech technologies

Speech Technology Center offers a wide range of its innovative proprietary speech processing technologies.

• Speech Enhancement and Noise Cancellation (Clear Voice) - proprietary, original state-of-the-art algorithm developed for automatically removing ambient noise from voice and other previously recorded or transmitted sound signals.
• Anti-Reverberation (Reverb Clipper) - new anti-reverberation technology. Increases the clarity of speech signals recorded indoors with reverberation time up to 3 seconds.
• Noise Robust Voice Command Recognition (VoiceCom) - algorithms of voice command recognition for SAPI. Language-independent and speaker-independent technology allows operating computer and other electronic devices by means of voice.
• Biometric Voice Verification (Voice Key) - voice-based access control algorithm performing user authentication by voice password. In just 0.5 sec Voice Key will identify a user and define his access rights to information and services.
• Biometric Voice Identification (VoiceNet SDK) - is the unique language-independent automatic speaker identification technology based on the fusion of several independent identification methods.
• Voice Alteration (Morpher) - is used to change the voice to hide the identity of a speaker. The technology is used in courts to protect witnesses.
• Speech Speed Changing (Sound Stretcher) - playback of speech at variable speed (fast/slow) without pitch distortion. Technology allows tuning the duration of advertisements or news reports not affecting the quality of speech.
• Music Spotting / Identification (Jingle Tracker) - original state-of-the-art algorithm developed to spot music samples in a sound stream. Jingle Tracker allows spotting radio advertisements, reports or music compositions while on-air.
• Audio Data Mining (Voice Digger) - is the technology for automatic keywords detection in speech databases or live speech.
• Speech Synthesis TTS (VitalVoice) - is a text-to-speech technology. This is innovative hybrid technology that combines Unit Selection and concatenative allophone approaches and guarantees attaining natural sounding synthesized voicing of any text. VitalVoice is the name for a whole family of TTS systems designed for different platforms and device types (PCs, smartphones, etc.). TTS systems of the VitalVoice family do sound not only natural but also emotional, according to the context and style.
• Voice Recognition STT (Russogragh) - is the first phoneme-based speech recognition technology for the Russian language. The technology is based on Hidden Markov Model (HMM) and supports MRCP v.1 & v.2 standards.