Research On Car Whistle Sound Recognition And Localization System Based On Microphone Array

In recent years,the number of cars in cities has been increasing with the development of economy.The car whistle noise population has become a serious problem.The impact of car whistle noise on people’s physical and mental health has increased.Therefore,the management of car whistle noise has received more and more attention,especially in the nearby roads of schools,hospitals and residential areas.It is a great necessary to apply microphone array technology to the process of car whistle supervision,and to realize the identification and localization of car whistle sound.This paper studied the car whistle sound recognition system and the car whistle sound source localization system based on the microphone array.First,this study built a sound recognition system.The system used BP neural network to identify the car whistle.The Mel coefficient characteristics,kurtosis characteristics,variance characteristics and geometric mean with count average ratio characteristics as the input characteristic parameters of BP neural network by analyzed the different characteristics of car whistle sound and other sounds in frequency spectrum.And the car whistle sound learning and recognition system can classify the sound of collected.The different whistle sounds characteristics were analyzed,and the whistle sounds are classified by the method of sub-frequency band energy feature extraction.The experimental results show that the recognition rate of the car whistle sound is above 94%.So the recognition system can identify the car whistle sound very well.Second,this study built a car whistle sound source localization system based on microphone array.Because the shape of microphone array has a great influence on the location accuracy of car whistle sound source,the study designed a four-channels spatial cruciform microphone array model.The localization system analyze the traditional sound source positioning algorithm,then improve a localization algorithm to locate the sound source of car whistle based on combination of time delay estimation and the spatial plane intersection method.The system uses STM32F407 controller to control the sound acquisition,processing,the photograph of whistling vehicles and data storage.The experimental results show that the distance error of car whistle sound source localization is controlled within 0.21 m.The system realize the location of the car whistle sound source accurately.The results also show that the error of the location also raise with increasing distance.Combined with the location results of whistle sound,the license plate characters of whistle cars are extracted and segmented.The license plate characters are identified and extracted completely through the template matching method.