Products

DSpotter, based on neural network model, support offline keyword spotting, performing high accuracy and better in noisy environments with low power consumption.
Support Always listening、Hands-free、Eyes-free.

Features

  • Phoneme-based speaker-independent (SI) command recognition
    • Creating voice commands immediately just with text input
    • No voice data collection, reducing development cost
  • NN-Based model
    • Long-distance speech recognition
    • Great performance in noisy environment.
  • Offline keyword spotting with low power consumption.
  • Supporting SOC、MCU and DSP platform. Ex: ARM CortexM series, CEVA TL3/TL4 series, Tensilica HiFi series,
    Synopsys ARC, Andes RISC-V, RX.

Languages

More than 26 languages: Mandarin, English, Japanese, Korean, Turkish, Vietnamese, German, French, Spanish,
Italian, Norwegian, Hindi, Russian, Cantonese...

4 languages/month

Solutions

Local trigger-word and voice command in smart speaker, TWS, car system, IoT and smart appliances.

CSpotter is a keyword spotting engine. It supports the functions of always-listening, keyword spotting and non-keyword filtering. Voice trigger does make hands-free control possible.

Features

  • Phoneme-based speaker-independent (SI) command recognition
    • Creating voice commands immediately just with text input
    • No voice data collection, reducing development cost
  • On-device user trained command (voice tag) and speaker verification (SV) capability
  • Supporting fix-point 32-bit MCU and 16/24-bit DSP

Solutions

Home media, navigation, or computer in sleeping mode etc.