NEC develops new technology for car voice noise removal

Recently, NEC has developed a noise removal technology that can significantly improve the voice operation accuracy of onboard systems.

In the age of automotive intelligence, voice operations have gradually become the standard configuration. However, when the air conditioner is used or the window is opened, the voice operation is not smoothly performed because of the influence of the surrounding noise. The in-vehicle speech denoising technology developed by NEC uses two microphones to absorb sound sources, and then performs noise processing through two steps, enabling the sound model to make the device easier to identify. According to reports, with this technology, correct voice operations can be accomplished even in a noisy environment that is no more than 5 times the original noise. Whether it is air conditioning or window opening, voice operations can be successfully completed.

So what is the specific processing step for this technology?

First, by analyzing the position of a noise source such as an air conditioner and the manner in which sound is propagated to the in-car microphone, the two microphones are placed one behind the other at the optimal position in the car with a small pitch. In this way, the noise from the front of the car and the speaker's voice can be accurately identified, thereby improving the efficiency of noise removal.

After that, noise is detected twice for the detected speech and noise. First, using the similarity of the waveform of the noise entering the two microphones, accurately capturing the waveform of the noise that changes with time, extracting the main noise components, and then using the speech correlation in the two microphones to further eliminate the noise, and through such processing, large Some noise has been removed.

Finally, due to the de-noising process, speech distortion may exist, which may make it difficult to identify the on-board system. In response, NEC has uniquely developed two speech models to make adjustments. Specifically speaking, the speech components and their changes are first modeled, and then the distorted speech is automatically selected in the two models and finally adjusted to easily recognizable speech.

NEC has always been keen on the research of voice technology in its efforts to build smart cities and promote social solutions. Prior to that, NEC has developed world-leading technologies such as speech recognition and noise removal. In the future, NEC will also consider research on speech denoising. Apply to more scenes such as shops and factories.

NEC's "Mid-term Business Plan 2015" is to achieve a safe, secure, efficient, and fair social environment for human's affluent life. NEC uses ICT technology to provide intelligent social infrastructure solutions and shoulders the "social value creation enterprise". Responsibility, to contribute to solve various social issues.