IP for mixed-signal SoCs enables voice activation with “WhisperTrigger”

March 10, 2015 // By Graham Prophet
Voice activity detection is considered as a must-have for a growing range of applications such as wearable devices or smart home appliances, according to Dolphin Integration, which has introduced the stand-alone, autonomous WhisperTrigger function.

Whereas the voice detection function is usually handled by a DSP or a voice processor, it now becomes possible to leave the rest of the SoC in stand-by or sleep mode while leaving only the WhisperTrigger in always-on mode. It can wake-up the system without voice data loss.

Depending on the microphone in use, either digital or analogue, two triggers are available - WhisperTrigger D and A.

The WhisperTriggers, designed as hard Silicon IP, are suited for triggering voice recognition algorithms thanks to their low detection latency enabling faster wake-up of the voice Processor.

Features of WhisperTrigger-D for a digital MEMS microphone include;

Saving up to 30% of DSP activity due to an efficient detection algorithm, avoiding numerous false detections compared to power-based algorithms

Ultra-low power consumption: only 15 μW added to codec stand-by mode consumption for Voice Detection

Detection Latency (DL): down to 8 msec depending on the type of phoneme

Noise Detected as Speech Rate (NDSR): 8% at 10 dB of Power Ratio (PwR) threshold level.

WhisperTrigger is embedded in several products in Dolphin’s portfolio:

sCODa96-H1-LB-IO-VD.01 (28 nm)

sCODS100-LB-IO-N.12 (40 nm)

sCODp-MT1-VD.02 (130 nm)

Dolphin Integration (France); www.dolphin-ip.com