Vehicle type identification based on engine sound and whistle

• Main Authors: LIN, WEI-YU, 林威宇
• Other Authors: Kuang-Yow Lian
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/3dq69f

碩士 === 國立臺北科技大學 === 電機工程系 === 107 === The purpose of this thesis is to realize an instant identification system for a variety of vehicle types on the road by means of voice recognition technology. These vehicle types include whistling vehicles such as ambulances, fire engines and police cars, as well as cars, locomotives and buses that are not whistling vehicles. Most of the literature on general vehicle identification and whistle recognition uses more complex neural network algorithms, or images, for analysis, which require a large amount of computation. Hence, it is difficult to implement on embedded systems. In order to achieve the purpose of real-time, the algorithm of this thesis can reduce the amount of computation of the program and successfully complete the identification. On the road, ambulances, fire engines and police cars emit obvious whistling sounds. This research directly captures the characteristics of the vocal characteristics of the whistling vehicles in the time domain, and successfully identifies the ambulances, fire trucks and police cars. As to non-whistling vehicles such as cars, locomotives, buses and heavy machines, this paper uses the support vector machine through training as the basis of classification for the engine sound of different vehicles. To this end, we establish a small feature database and perform comparison study. In terms of features, we calculate the low frequency, intermediate frequency and high frequency energy as the features to train. After obtaining the training model, we use it to reach the recognition of various types for the non-whistling vehicles.