Shenzhen Voice Interaction CenterCurrent location: Home > Shenzhen Voice Interaction Center
- The main members are led by PhDs in the field of speech/audio processing in the United States and China, with rich experience in speech processing and enhancement, and more than 20 years of research industry experience.
1. Our advantage-compared with the solutions on the market
1. Reduce computing resources by up to 50%, and perform significantly better in unfavorable environments
2. Lower hardware design restrictions
3. The microphone of the communication application is allowed to be mismatched;
2. Voice processing-basic features:
1. Voice enhancement: voice activity detection, noise suppression, beam forming
2. Acoustic echo cancellation: dual-talk detection
Three, adaptive deep machine learning sound spectrum analysis
Rule-based statistical model + neural network model for knowledge learning
2. Rule-based statistical models can use statistical data
3. To suppress noise, eliminate echo, and focus on human voice
Fourth, the knowledge-based machine learning model understands audio streams as well as humans
1. It can better separate speech from noise
2. Separate the near-end voice from the echo
3. Speaker recognition can be performed
4. Keyword detection;
Fifth, provide fully adaptive AEC without parameter correction: even in the case of dual-talk, there is no need to adjust the parameters.
Applications of beamforming and blind source separation:
1. Accept the microphone distance from 30mm to 200mm
2. The number of microphones that can be handled ranges from a single microphone to an array of up to 8 microphones, when the number of microphones is not less than 3
3. Provide accurate DOA information;
Six, our headset call noise reduction products:
1. Embedded microphone array voice enhancement processing
2. The overall solution of smart speaker voice wake-up