Ship Sound Signal Recognition Based On MFCC And SVM

In recent years,due to the extensive application of the Internet of Things,big data analysis and artificial intelligence,intelligent ships have become an emerging hotspot of joint research and development in the global maritime industry.In this paper,the main goal is to autonomously recognize the meaning of the ship’s sound signal and reduce collision avoidance.In view of the characteristics of noisy background and many interfering signals of the ship’s sound signal,a ship’s signal recognition and analysis system is designed.Through the interpretation of the signal,other ship’s dynamics or intentions can be known,so as to achieve the purpose of information interaction between intelligent ships and traditional ships,and provide basic support for ship’s autonomous navigation.The main research contents are as follows:(1)Three noise reduction methods are adopted to analyze the collected ship sound signal with noise and four types of noise interference signals.Through experimental comparison,the improved spectrum subtraction method with stable noise reduction effect and small noise residue is selected as the denoising method in this paper,which reduces the risk of signal loss and can achieve better noise reduction effect.(2)The Mel-scale frequency cepstral coefficients(MFCC)of sound is extracted.By introducing second-order difference and wavelet packet decomposition,the MFCC is improved and new MFCC characteristic parameters are formed,which effectively improves the dynamic characteristics of MFCC and the characterization ability of high-frequency information.Combined with the time-frequency characteristics of sound,the 42-dimensional feature vector set is constructed,which can comprehensively represent the sound information and lay a foundation for improving the accuracy of ship sound signal recognition.(3)The advantages and disadvantages of the sound classification and recognition model were analyzed and studied.A ship sound signal classification and recognition model based on support vector machine(SVM)was designed.Through continuous optimization and testing of the model,the classification accuracy of the ship sound signal and interference signals of the model reached 91.33%,which further verified the effectiveness and accuracy of the recognition method.(4)Using the method of MFCC cepstrum distance to detect the end points of the ship sound signals,calculate the duration of each section of the sound signals,and output the meaning of the sound signals according to the established ship’s sound signals analysis database.With the help of MATLAB,the ship sound identification and analysis system can is built,so that users can intuitively and quickly complete the identification and analysis of ship sound.