Road sound recognition method and system, computer equipment and readable storage medium

Abstract

The invention discloses a road sound recognition method and system, computer equipment and a readable storage. The road sound recognition method comprises the steps that data samples and sample categories of road sounds areacquired; time-frequency decomposition, frequency conversion, logarithmic operation and derivation areperformed on the data samples in sequence to obtain logarithmic Mel characteristics of the data samples; the logarithmic Mel characteristics areinput into aconvolution cyclic network model for training according to the sample category until the convolution cyclic network model meets a preset training end condition; and the convolution cyclic network model after training isutilized to identify and classify the voice data to be processed. According to the road sound recognition method and system,the computer equipment and the readable storage, the logarithmic Mel characteristics of the road sound data samples are taken as inputs of the convolution cyclic network model,and the model which can be used for identifying multiple sounds in a complex traffic scene is trained, so that the accuracy of road traffic event detection is improved.

Description

technical field [0001] The invention relates to the technical field of voice recognition, in particular to a road voice recognition method, system, computer equipment and a readable storage medium. Background technique [0002] With the development of science and technology, intelligent transportation has gradually become an important means of road monitoring. At the same time, in order to make road monitoring more intelligent, traffic incident detection technology has emerged as the times require. Traditional road traffic incident detection technology mainly relies on video detection. However, video detection is directional, and it is difficult to complete correct recognition in bad weather, poor lighting conditions, lens pollution, etc., and the accuracy of detection cannot be guaranteed. [0003] The inventors have found that road sounds carry a large amount of event information. At the same time, multiple sounds may appear on the road at the same time, such as car whistl...

AI Technical Summary

Problems solved by technology

At the same time, the inventors have found that most of the existing research on sound recognition is limited to identifying the most prominent event information at the same time, such as laughter, applause, etc., while other event information is lost, which obviously does not meet the requirements of road sounds. Requirements for voice recognition in complex scenes

Images