Artificial intelligence voice module

The brain neural network processor core (BNPU) equipped with the company's independent intellectual property rights can perfectly support the high-performance DNN computing integrated application processor core and various common control interfaces. It can be used to design various intelligent voice product solutions to achieve local high precision. Multi-lingual, large vocabulary speech recognition and control functions.

Working power consumption: 0.3w

Speech recognition rate: 95%

Recognition distance: 10m

Response speed: 0.5s~0.8s



AI neural network processing algorithm

The hardware processor based on the deep AI neural network (DNN) enables the collection, calculation and decision-making of voice data to be performed locally, without relying on the network, and with small delay, which can protect user privacy.




Speech Recognition

Fast and accurate conversion of voice into corresponding text, far-field recognition rate of more than 95%, response time of up to 0.3S, support for Chinese, English, and some dialect recognition.



Voiceprint recognition

Listen to the voice, automatically extract the speaker's voice characteristics, identify the identity.



Natural language processing

Identify natural language and semantics and respond quickly.



Wheat array noise suppression

Support classic four-grain ring, linear array, 360-degree omnidirectional sound collection, can suppress ambient noise outside the target voice direction, and ensure the accuracy of noisy environment speech recognition.




Echo cancellation

The voice wake-up function is realized, and the echo suppression ratio is over 20dB, which can be used for products such as story machines and smart speakers.