

Full link
design

Scene noise
reduction

AI skill
customized

Wakeup words
customized

Scene skills

Far-field
interaction
Core strengths

Multi-language support
​>97%
Cloud recognition rate
<0.3 s
Identification time
≈0.3 s
Dynamic compilation
is time-consuming
​>94%
3m far-field
recognition rate

Noise-resistant
far-field recognition
In-depth training of noise reduction models for household noise.
Excellent noise reduction for common household noises such as fans and air conditioners.

Cloud-based
recipe functions
Supports a large
number of recipes in
the cloud for timely
use.

On/Off-line Voice
Recognition
Support on/offline
voice recognition, fast
offline command response, rich online.

Device IoT
Provides Wi-Fi API
interface for fast IoT

Voice SDK

Based on the DUI open platform, it
supports the output of voice single-point technology, including speech recognition SDK, speech synthesis SDK, natural language understanding SDK/AI skill call, etc.

Total Hardware solution

Built-in 2/4 microphone array, reliable
power management, touch, screen and
other interfaces.

Function customization

Wakeup words customization, off-line
command words customization.
Typical application module solution

Off-line voice recognition module
-
2-mic linear array
-
Wake-up words customization
-
Nearby wake-up
-
Interface: UART, I2S
-
3~5 meters range
-
Support 50 off-line commands

Off-line speech recognition module
-
2-mic linear array
-
Support Wi-Fi to cloud ASR
-
Wake-up words customization
-
Nearby wake-up
-
Interface: UART, I2S
-
3~5m recognition distance
-
Full duplex
-
Support 50 off-line commands
-
Support OTA

On-line core module
-
Support 2 MIC array
-
Support Wi-Fi to cloud ASR
-
Wake-up words customization
-
Nearby wake-up
-
Interface: UART, I2S
-
3~5 meters range
-
Full duplex
-
Support 50 off-line commands
-
Support OTA
-
Stamp hole design

Off-line core modules
-
Support 2 4 MIC array
-
Wake-up words customization
-
Nearby wake-up
-
Interface: UART, I2S
-
3~5 meters range
-
Full duplex
-
Support 50 off-line commands
