Phonexia Voice Activity Detection (VAD) speech technology detects which parts of audio recordings contain spoken content.
Cut through large amounts of audio files quickly with the automatic detection of whether they contain speech or not, and where exactly such content appears in them.
Run advanced speech and voice analyses efficiently and save on hardware costs by making sure that only relevant audio files containing speech content are analyzed.
Provide chatbots with the ability to know whether there is a someone speaking to them or not so they can respond to voice commands in a natural and timely manner.
Technology Details
• Trained with an emphasis on spontaneous telephone conversations
• Is language-, accent-, text-, and channel-independent
• Applies state-of-the-art channel compensation techniques compatible with the broadest range of audio sources possible: GSM/CDMA, 3G, VoIP, landlines, satellite phones, etc.
WAV or RAW (PCM unsigned 8 or 16 bits, IEEE float 32-bit, A-law or Mu-law, ADPCM), FLAC, OPUS; 8 kHz+ sampling (other audio formats are automatically converted).
• XML/JSON format with all results
• Results files with labels (speech vs. non-speech segments)
Detects voice activity in any language as the technology is language independent.
Approximately 150x faster than real-time processing on 1 CPU core—for example, a standard 8 CPU core server processes 28,800 hours of audio in one day of computing time.
Schedule a demo with our experts and see for yourself how Phonexia Voice Activity Detection technology can help your business.
We will show you a free online demo tailored to your needs, showcasing Phonexia speech and voice recognition technologies and the ways you can use them to achieve your project’s goals.
