Voice AI development & deployment for robots
VoxEdge AI is a full-stack edge voice-AI partner for robotics companies. We build and deploy custom on-device speech recognition (ASR), wake words, text-to-speech (TTS) and dialogue logic on your edge silicon, plus the full acoustic front-end: microphone arrays, beamforming, echo cancellation and noise suppression.
Core technology
- Microphone array pickup — 4-mic linear and 6-mic circular arrays, 5m+ far-field
- Sound source localization (DOA) and beamforming for directional pickup
- Echo cancellation (AEC) and AI noise suppression (SNR +15dB)
- Voice activity detection (VAD), accuracy ≥90%
- Speech recognition (ASR) — >98% far-field accuracy, <300ms latency, custom hotwords
- Large language models (LLM) with emotion detection and multi-turn memory
- Text-to-speech (TTS) and voice cloning, MOS >4.0
Applications
- Commercial service robots — far-field pickup in malls, airports and large spaces
- Quadruped robot dogs — acoustic tuning for motion and motor noise
- Industrial & specialized robots — high-noise, high-safety voice control
- Humanoid robots, companion robots, AI toys
Edge silicon
NVIDIA Jetson, Qualcomm Robotics, Rockchip, Google Edge TPU and custom NPUs — TensorRT, RKNN, SNPE, TFLite, ONNX Runtime.
Contact
Email: kaixinshier@gmail.com — read more on our blog and services page.