Research On Vehicle Interior Speech Intelligibility Evaluation And Active Control Method

Automotive interior noise can affect the performance of ride comfort, hearing damage,speech intelligibility. It also deprived occupant of identifying the wanted information fromvarious sound signals resulted from inside and outside the vehicle. The interior noise levelhas become one of the important performance indicators of ride comfort, and it directlyaffects the competitiveness and consumer orientation towards to the automotive products.The A-weighted sound pressure level (SPL) is almost used as a vehicle noise levelevaluation among current automotive noise regulations and standards. These laws andregulations set a limitation to SPL of automotive noise, but did not take the interior noisecharacteristics into considered in depth. The quicker development of the automotiveindustry is the higher requirements are put forward to automotive noise and sound quality.Sound quality evaluation has been developed rapidly because single A-weighted SPL hasbeen unable to fully meet task in automotive industry. It pointed out that people feel thenoise affected by human physiological and psychological, and reflect a subjective feelingand evaluation. Supported by National Natural Science Foundation project,“the analysisand evaluation of sound quality of interior noise and adaptive active control methods”(itemnumber:50975119), automotive sound quality are conducted to analyze the characteristicsof interior noise and the methods of adaptive active control are researched.Firstly, the test of speech intelligibility is carried out under interior noise when car isoperating at different conditions. The relationship between speech intelligibility andacoustics objective evaluation indexes is studied. And then the establishment of an objectiveevaluation model of speech intelligibility is consistent with the test results. The testingresults of speech intelligibility are obtained thought organizing many audients to do the subjective evaluation of the speech intelligibility in the vehicle interior noise background.At the same time, the acoustic objective evaluation indexes of the vehicle noise sampleshave been calculated, such as speech intelligibility index, loudness, sharpness, roughness,2different weighted SPL, PSIL, SIL, and SIL(3). Two objective evaluation models of speechintelligibility with the independent variable of acoustic objective index are established byregression analysis. These models are consistent with the testing result. The advantage ofmultiple linier regression and quadratic polynomial model has been compared. The resultshows that the quadratic polynomial objective evaluation model of speech intelligibility,related the SII, has better fitness and signification level.Secondly, the vehicle interior noise frequency band has been divided according to ERB,and the different special bandwidth noise has been processed with parametric filteringmethod and the processed noise samples are used to analysis the special bandwidthsensitivity of acoustic objective evaluation indexes. In addition, acoustic objectiveevaluation index of the all noise sample has been calculated, and the sensitivity of the SPLchange of different special bandwidth noise on the acoustic objective evaluation index hasbeen analyzed. And the special bandwidth sensitivity of speech intelligibility has beentechnically analyzed. The sensitive frequency bandwidth of speech intelligibility index isconcentrated at the5th and nearby special bandwidth and its relative gradient curve isfunnel-shaped. The change of the SPL in these special bandwidths can significantly adjustthe value of speech intelligibility index. The sensitive bandwidth of the psychologicalobjective evaluation indexes such as loudness, sharpness, roughness and A-weighted SPLare concentrated in the first5special bandwidths. So the change of the SPL in these specialbandwidths has been a dramatic influence on these indexes.Furthermore, the mid-low frequency band character of the vehicle interior noise hasbeen analyzed to acquire the optimal target of speech intelligibility. The speechintelligibility index and sound quality objective evaluating model’s score has been proposedas testing indexes, and a testing schedule with6special bandwidths and5testing levels hasbeen designed to evaluate the question of interior noise optimization. Do the orthogonally experiments on the interior noise sample of the working condition of60km/h,100km/h and120km/h, and the optimal combination of the optimal levels in each special bandwidthbased on the speech intelligibility and sound quality evaluating model has been obtainedthought the range analysis. The all results are validated to improve the sound quality byexperiments based on the conclusion from the combination of orthogonal experiments.Then, an ANE active control algorithm with the FSLLMS core is proposed. It canachieve selective counteract or supplement to the different special bandwidths’ SPL. Thespeech intelligibility adaptive active control model is established in MATLAB/Simulink. Dothe active control simulation on the interior speech intelligibility and sound qualityimprovement using interior active control system with the working condition of100km/h.To improve interior speech intelligibility, the target is to enable the objective evaluatingmodel reach optimal active control level which can effectively reduce the noise SPL andincrease the speech intelligibility index by14.3%. To improve the interior sound quality, thetarget is enable the objective evaluating model’s score reach the optimal strategy which canmake better sound quality comprehensive performance with improvement rate of8.93%.Finally, an adaptive active controller with proprietary intelligent property of the speechintelligibility is developed and the relative control program is written and the adaptiveactive control system of speech intelligibility is constructed. Do the real vehicle noisetesting in the working condition of60km/h,80km/h and120km/h, and analyze thedifference between the interior SPL before and after control. The adaptive active controlsystem of speech intelligibility is aimed to improve the performance of speech intelligibilitywhich can reduced the value of loudness with maximum reduction of7.14sones andenhance speech intelligibility with the maximum improvement rate of17.05%; the system isaimed to effectively improve the score of sound quality objective evaluation model whichcan achieve the biggest improvement rate of14.48%. Besides, the interior speechintelligibility index was also increased by14.96%. The satisfied test results can present that the proposed algorithm and corresponding control system is not only effective but also canachieve improvement in interior sound quality.