中文
中文  /   English
  • SKWS key words spotting solution

    SKWS key words spotting solution runs independently on the local end of the device and responds to the user's voice command in real time. The combination of microphone array and SVE solution can achieve accurate speech recognition in noisy environments. Combined with the voice sensor, SKWS solution can accurately detect the voice endpoint and confirm the user, reducing the standby power consumption and false recognition rate. The solution supports parallel recognition of multiple control commands, without cloud confirmation, so that user commands can be responded instantly, which can improve the experience.

Technical composition & specification
  • Support common command words(only Chinese currently)

    Previous Song/Next Song
    Volume up/volume down
    Play music/pause/resume playback
    Command words can be customized on demand

  • Noise suppression

    Average recognition rate in noise scenario (5 dB SNR condition) > 85%
    Average standby power consumption ~1 mA (ARM Cortex M4F)

  • Voice Activity Detection (VAD)

    Used to determine whether the voice was made by the user
    Remove signal interference caused by non-speech activities such as chewing and walking

Mass production cases
  • Xiaomi Air 2/2s
  • Wyzeband