Voice Interface Voice Interface Module

Dual-Microphone Voice Array

XMOS XVF3610-based dual-microphone voice interface with 5-meter far-field pickup and dual-channel ASR/voice call output.

5-meter Far-Field Pickup ASR & Voice Call Plug and Play

Contact Sales

Instantly Optimized Voice

The compact dual-microphone array with audio front-end processing technology clearly captures voice within a 5-meter range. The powerful MCU can simultaneously process and optimize both speech recognition and conference calls through dedicated dual-channel outputs.

Technical Specifications

Microphone Count	2 PDM microphones
Far-Field Pickup Range	Up to 5 m
Sample Rate	48 kHz
Audio Interface	USB Audio Class 1.0 / I²S
I²S Mode	Master / Slave
Output Channels	Dual — ASR + Voice Call
DSP Algorithms	AEC, Noise Suppression, AGC, Beamforming, VAD
End-to-End Latency	< 10 ms
Firmware Storage	External QSPI Flash (plug-and-play upgrades)
Host Compatibility	Linux, Android, MCU

Key Features

All-in-One DSP Algorithms

With built-in Acoustic Echo Cancellation (AEC), interference canceller, noise suppressor, and Automatic Gain Control (AGC), our product can clearly capture far-field voice signals up to 5 meters without loading the host CPU, effectively eliminating echo and background noise. This reduces debugging workload by up to 80%, and its barge-in feature makes conversation flow more efficient and coherent.

Dual-Channel Output • ASR & Voice Call

One channel preserves richer voice spectrum details specifically for ASR recognition; the other channel optimizes voice call clarity. Both channels operate simultaneously, perfectly supporting products that require ASR functionality.

Fast Integration, Accelerate Time-to-Market

Thanks to its compact size, you can easily design acoustic structures for this product and integrate it into various small and medium-sized devices.

Multiple Interfaces • Flexible Integration

Supports USB Audio Class 1.0 and master/slave I²S, easily interfacing with Linux, Android, or MCU hosts; firmware stored in external QSPI Flash enables plug-and-play upgrades.

Application Scenarios

Service Robots

On-device far-field voice enables robots to accurately understand commands within 5 meters, helping hotels, malls, and factories achieve safer hands-free operation.

Smart Home

Clear ASR signals ensure accurate command recognition, allowing you to easily add voice control to smart home gateways and smart TVs.

Online Meetings

Provides clear voice quality for video conferencing, integrable into webcams and interactive whiteboards.

Meeting Transcription

Optimized audio output supports high-accuracy speech recognition, making meeting records and AI-assisted meeting notes more accurate.

Customer Case Studies

NDA Protected

Smart Livestock

Intelligent Livestock Health Monitoring System (Embedded)

Low-cost, high-precision health alerts for livestock farms

A global leader in livestock management (NDA) needed to identify and locate abnormal acoustic events (such as coughing) in noisy barn environments at low cost using on-device real-time inference. Pawpaw provided an embedded solution: lightweight on-device AI inference combined with array beamforming, noise suppression, and DOA source localization to achieve precise, low-latency health alerts, along with OTA update support for safe incremental rollout.

Integrates microphone array and AI inference on a single chip for local real-time alerts. Replaces multi-chip architecture in cost-constrained scenarios with support for OTA algorithm updates.

Key Features:

On-device inference: End-side latency reduced to milliseconds, accuracy maintained above 90% of baseline
Array signal processing: Improved SNR via beamforming and noise suppression
DOA source localization: Pinpoints specific coughing events in noisy environments
Model quantization/pruning: Ensures smooth operation on embedded chips
OTA: Continuously iterate and rapidly deploy new algorithms via OTA

AIoT Platform

Microsoft Azure: Microphone Array Development Kit

The Microsoft Azure team, while building a rapid development platform for edge AI applications, sought to integrate a low-power, real-time microphone array module fully compatible with their API to simplify audio processing for developers.

Built a custom microphone array solution for the Microsoft Azure AIoT platform. USB stack optimization enables seamless integration with Azure APIs, allowing developers to access low-latency, stable audio input without writing embedded code.

Key Features:

Deep hardware and software customization: Ensures high compatibility between microphone array and Azure dev kit
USB stack optimization: Stable and efficient data transfer between host and development board
Iterative collaborative development: Close cooperation with Azure team to validate and optimize each hardware revision
Low latency and high stability: Meets edge AI requirements for real-time processing and reliability
Zero developer overhead: Call audio input functions without embedded programming