离线语音识别

Offline Speech Recognition

No internet or app download required for high-accuracy speech recognition. Based on full-stack deep learning DNN algorithms, supports offline natural speech (offline NLP), multi-intent understanding, command word self-learning, etc. Covers multiple languages including Chinese (Mandarin/dialects), English, Japanese, Korean, Arabic, Vietnamese, and more.

Excellent recognition experience

Recognition rate >97%, response time <0.2s 

Powerful features

Supports deep noise reduction, offline natural speech, multi-intent understanding, etc. 

Easy to implement

Uses serial communication for rapid development and verification 

Mature solution

Reduces development cycle and support quick product landing 

Single-Mic Speech Recognition

Implements offline speech recognition with a single microphone, supporting offline natural speech and fixed phrase recognition; also supports multi-intent understanding, acoustic echo cancellation, command word self-learning, and multiple wake word switching.

Offline natural speech

Multi-intent understanding

Single-Mic Speech Recognition

The voice chip converts captured voice into commands and communicates with the MCU of the control board or other devices through a serial port. After receiving the commands, the MCU or device responds accordingly, or sends commands to the voice chip for voice announcements. SDK and documentation are provided for advanced development. This solution has been mass-produced in various fields such as fans, electronic clothes hangers, heating tables, tea bars, and lighting.

Chip Datasheet



Module Dadasheet



Dev Board Manual



SDK



Single-Mic Speech Recognition

Single-Mic Deep Noise Reduction Speech Recognition

Uses intelligent algorithms based on deep neural networks to filter complex background noise and improve recognition effectiveness. Can be used for voice recognition interaction in environments with noise levels above 70dB, such as electronic clothes hangers and range hoods.

Deep noise reduction

Single-Mic Deep Noise Reduction Speech Recognition

The voice chip converts captured voice into commands and communicates with the MCU of the control board or other devices through a serial port. After receiving the commands, the MCU or device responds accordingly, or sends commands to the voice chip for voice announcements. SDK and documentation are provided for secondary development. (Joint training of noise reduction and speech recognition models is required for specific devices)

Chip Datasheet



Module Dadasheet



Dev Board Manual



SDK



Dual-Mic Speech Recognition

Supports offline natural speech and fixed phrase recognition; supports multi-intent understanding, acoustic echo cancellation, command word self-learning, and multiple wake word switching. Also supports algorithms such as dereverberation, deep source separation, and sound source localization, and various algorithm combinations to meet different application scenarios.

Offline natural speech

Multi-intent understanding

Sound source localization

Support multiple algorithm combinations

Dual-Mic Speech Recognition

This solution adopts Dual-Mic，The voice chip converts captured voice into commands and communicates with the MCU of the control board or other devices through a serial port. After receiving the commands, the MCU or device responds accordingly, or sends commands to the voice chip for voice announcements. SDK and documentation are provided for advanced development. This solution has been mass-produced in various fields such as fans and lighting.

Chip Datasheet



Dev Board Manual



Single-Mic Voice Controller

The voice chip is used for both voice recognition and device control. The chip responds to or controls devices through I/O based on voice command. Supports offline natural speech, fixed phrase recognition, multi-intent understanding, acoustic echo cancellation, command word self-learning, and multiple wake word switching.

Offline natural speech

Multi-intent understanding

Device control

Single-Mic Voice Controller

Our chip has more than 20 I/Os for controlling various external devices. Our chip can directly replace the original MCU for related external device control, which is more cost-effective and compact compared to serial port and control communication solutions.

Chip Datasheet



Module Dadasheet



Dev Board Manual



Single-Mic Inteligent voice & IR Control

The voice chip converts captured voice into commands and sends corresponding IR control codes to control devices. Supports 98% of air conditioners IR code and can acquire new IR code through advanced development; supports mini program remote control. Supports offline natural speech, fixed phrase recognition, multi-intent understanding, command word self-learning, and multiple wake word switching.

Offline natural speech