Kopin Audio Chip Hears WhispersMarch 7, 2017
Author: Mike Demler
Kopin’s Whisper is a tiny low-power speech processor that integrates a proprietary DSP the company calls a voice-extraction filter (VEF). The chip runs algorithms that enable it to separate voice commands as well as a stream of spoken words from extremely noisy backgrounds, such as in a factory, a crowded room, or a busy road with sirens and traffic. Whisper primarily targets wearable devices such as smart eyewear.
Aurisound originally developed Whisper’s voice-extraction technology, but Kopin acquired that startup in 2013. The chip serves as a voice-processing front end in an automatic-speech-recognition (ASR) system. It runs the DSP algorithms to extract speech data, which it transmits to a host though one of its serial interfaces. The host will either run speech-recognition software locally or transmit the data to a cloud-based service such as Amazon’s Alexa Voice Service (AVS).
Along with its low power, Whisper’s main differentiator is its ability to clearly process speech in environments that the American Speech-Language-Hearing Association (ASHA) classifies as extremely loud. In a demonstration that included noise simulating a crowded room, we could barely hear the speaker’s voice, but speech-recognition software running on a PC correctly translated every word.
Subscribers can view the full article in the Microprocessor Report.
Subscribe to the Microprocessor Report and always get the full story!