Naive Bayesian classifier target recognition based on ridge features

Abstract

Traditional time-frequency algorithms do not have a high feature recognition rate for targets in a low signal-to-noise background, and the number of operations for classification using plain Bayes is large and slow. To solve the above problems and achieve recognition of different ship echo signals in water, this paper proposes a target recognition method based on time-frequency ridge features by using collected audio data of different ship types, combined with Fourier Simultaneous Compression Transform (abbreviation: FSST), starting from time-frequency features, by extracting the FSST time-frequency ridge and finally using the Hopper Parsimonious Bayes classifier to obtain respectively. The recognition rate of ships under FSST and the recognition rate of ships based on ridge line features are obtained respectively. The final comparison shows that the correct recognition rate of the plain Bayesian classifier using the spine features of the time-frequency map is much higher than that of the conventional FSST time-frequency method, and it also outperforms the FSST in terms of computational power and speed.