Background

In the rapid development of modern smart devices and Internet of Things (IoT) applications, voice interaction has become a crucial part of user experience. Users' expectations for smart devices are not limited to basic voice command recognition but also include accurate recognition from a distance, multi-language support, and stable performance in noisy environments. To meet these demands, advanced DSP technology and efficient keyword spotting (KWS) solutions become essential.

Solution Overview

PXU316-Voice is a solution that combines dual-microphone front-end audio processing (Audio Front-End, AFE) with a Keyword Spotting (KWS) AI model. Thanks to its high-performance dual-microphone AFE algorithm, it can support wake-up commands within a range of 5 meters and also supports barge-in functionality. PXU316-Voice can recognize multiple languages through local AI models, capable of distinguishing up to 100 voice commands. Additionally, it can provide optimized clear voice signals for speech recognition (ASR) in user products via USB Type-C.

Features

Far-field voice pickup

Achieve 360° far-field audio pickup within 5 meters using a dual-mic algorithm.

Multilingual Keyword Spotting

The local offline KWS model supports up to 21 languages, with an accuracy of up to 95%, and the command set can be customized.

Application

The PXU316-Voice can be used in various intelligent voice interaction devices, including but not limited to smart home devices, smart speakers, smart appliances, and other consumer electronics. It can also be applied in smart healthcare, intelligent driving systems, and other fields.

KWS Voice Control Example for Smart Interactive Whiteboard