In the noisy background sound of places such as airports and train stations, the microphone will collect the ambient sound together, and will be greatly disturbed when answering, so it is not easy to hear each other's voice clearly. With the emergence of noise-reducing microphones for smartphones, a supplier of microphone noise-reduction chips, the noise in the conversation becomes smaller and smaller.
Microphone noise reduction chip supplier smartphone noise reduction microphones are currently the more widespread noise reduction method of dual microphone noise reduction method, but dual microphone noise reduction itself has certain limitations, because the algorithm needs to run on the mobile terminal, so the scale cannot be too large. The effect is relatively limited. When no longer speaking, even if it is close to the mouth, the calculation of the human voice peeling will produce an intermittent situation, and the microphone is actually collecting environmental noise. And when users use Bluetooth headsets or wired headsets to make calls, the dual-microphone noise reduction algorithm on the mobile phone often fails.
With the increase in computing power of microphone noise reduction chip suppliers and the continuous optimization of AI algorithms, microphone noise reduction chip suppliers have established a data set containing noise + human voice and pure human voice, and trained it with a deep neural network. A "speech enhancement regression" algorithm that separates noise and human voice. The noise reduction capability of this algorithm is not inferior to the dual microphone mode from the application point of view, and it can even solve the problem of intermittent noise that cannot be overcome by dual microphones. The previous dual microphone mode will cause occasional intermittent noise when people answer the phone when they are moving. It is still difficult to capture, but the speech enhancement regression algorithm can separate the human voice from it.
At the same time, the update iteration of the microphone noise reduction chip supplier's algorithm is more agile and efficient than the hardware, and the user experience can be greatly improved. Some innovative features can come to users more quickly and get feedback. Microphone noise reduction chip suppliers have a broader imagination for their terminal products in terms of R&D cost and product form. Especially after the future 5G application, the algorithm model of microphone noise reduction chip suppliers may no longer need to stick to the terminal computing power, and it may change the way users use terminal products.