Research On The Method Of The Driver's Speech Enhancement

With the development of automotive electronics,in-vehicle electronic equipment has become increasingly popular.And the traditional control methods,such as buttons and knobs,are gradually being replaced by the control method of voice.This method can not only improve the driving experience,but also improve the driving safety of the driver due to its non-distracting characteristics.However,there will be a lot of noise in the actual driving environment,which will not only affect the reception quality of the driver's voice signal,reduce the recognition efficiency of the vehicle's voice interaction system,what's worse it may also affect the voice control function of the voice interaction system,making the invehicle electronic equipment unable to work well.Therefore,enhancing the driver's voice to improve the quality and intelligibility of the driver's voice signal has become an urgent problem to be solved.This paper first researches the evaluation indicators of the speech enhancement algorithm in the car,and compares the evaluation indicators of the signal-to-noise ratio,the Perceptual Evaluation of Speech Quality(PESQ)score and the Speech Intelligibility Index(SII),the simulation results show that the SII is more suitable for evaluating algorithm performance from the perspective of psycho-acoustics.Secondly,in order to improve the quality of the driver's voice signal collection,the boundary element method was used to simulate the sound field in the car to optimize the placement of the microphone array.At the same time,a voice signal collection system that based on the microphone array was built,and the hardware and software of the system were designed to realize the real-time collection and storage of the driver's voice signals.Finally,a reasonable scheme of the driver's speech enhancement that combines the minimum variance distortion-free beam-forming and the SPSIN speech intelligibility improvement algorithms was designed to implement the enhancement of the driver's speech signal,and the evaluation index was used to evaluate the algorithm performance.This paper uses the uniform linear microphone array to collect and enhance the driver's voice signals at a uniform speed of 20 km / h,40 km / h,60 km / h and 80 km / h.In the experiments of the driver's speech enhancement under the different speeds,the MVDRSPSIN algorithm proposed in this paper improves the PESQ score of MVDR beam-forming by an average of 77.9% and a maximum of 96.1%,and the SII by an averages of 34.78% and a maximum of 75.9%.And at the same time,the signal-to-noise ratios of the MVDRSPSIN algorithm and the MVDR beam-forming algorithm are the same,which means that with the improvement of clarity,the MVDR-SPSIN algorithm does not reduce the signal-tonoise ratio of speech signals,and the experimental results show that the method proposed in this paper can achieve the improvement of the quality and intelligibility of the driver's voice signal under the influence of the noise.Through the research of the performance evaluation index of the speech enhancement algorithm,the distribution of the sound field in the car,the sound collection system and the speech enhancement algorithm,the driver's voice signal has been enhanced,thereby further improving the driver's driving safety and driving experience.It not only provides a powerful reference solution for the driver's speech enhancement,but also makes a great significance to the development of the automotive industry.