An Innovative Approach for Vehicle Speed Recognition through Noise Emission Analysis by Using Machine Learning

This paper presents a comprehensive methodology for identifying vehicle speed ranges based on the characteristics of vehicle noise emissions. In the process of speed recognition, we introduce the innovative SoundSpeedBi-GRU model, which effectively integrates critical components such as frequency selection, feature extraction, and speed classification of vehicle noise. During the feature extraction phase, we conduct a detailed analysis of the distribution characteristics of frequency features across different speed intervals to identify the most significant intervals for further exploration. Through rigorous comparative analysis, we establish the most effective method for noise feature extraction, enabling the precise extraction of key features that reflect vehicle speed information. Utilizing a bidirectional gated recurrent unit (Bi-GRU) for speed classification, our approach demonstrated superior classification accuracy, ultimately achieving a noteworthy accuracy rate of $\mathbf{9 6. 9 3 \%}$. These results underscore the potential of the SoundSpeedBi-GRU model in the domain of vehicle speed recognition, offering an efficient and reliable solution for speed classification based on noise characteristics.