Research On Detection And Recognition Technology Of Cetacean Call

In the marine environment,whether it is humans or marine organisms,most situations use sound waves to communicate.Cetaceans use calls to communicate within and between populations,and cetaceans have the ability of echolocation.They use sound waves to detect and locate objects.It is worth studying the biological sonar and communication mode of cetaceans,which plays a role in promoting the development of underwater detection and positioning and underwater communication technology.Nowadays,cetaceans are becoming increasingly endangered,and protecting cetaceans is an important task to maintain the marine ecological environment.However,the marine environment is complicated,and it is obviously difficult and inaccurate to observe the distribution of cetaceans by vision.Therefore,passive acoustic monitoring is usually used to monitor the current distribution of cetaceans in the sea by detecting and identifying cetaceans in audio samples collected in the sea.The detection and recognition of cetaceans' call signals is of great significance to the general survey of the distribution of cetaceans and the protection of cetaceans.In this paper,cetacean call signal as the research object,the detection and recognition methods of cetacean call signal in different types and different applications are studied,which lays a foundation for understanding the cetacean call,the general survey of cetaceans and the protection of cetaceans.The main contents of this paper are as follows1.A method based on kurtosis for detection and recognition of cetacean click signals is studied.The kurtosis value of each frequency band is calculated as the detection quantity,and the kurtosis value under the condition of "noise only" is fitted as the detection threshold,and the detection results are obtained by comparison.This method can solve the problem that the energy detector only works under high signal-to-noise ratio(SNR).The kurtosis tick detector can guarantee the detection effect under the condition of weak signal-to-noise ratio,and can detect the bandwidth of the click.2.A cetacean whistle detection and recognition method based on the known acoustic features is studied.The template is made by the known features,and the similarity between the sample to be detected and the template is calculated by template matching method as the feature vector to make the sample set,and then the gradient lifting decision tree is used as the classifier for detection and recognition.This method can provide excellent detection rate for the call detection with known features.3.To study a detection and recognition method of cetacean whistle sound for detecting unknown acoustic characteristics.Through signal processing method,the signal is converted into a time-frequency image through short-time Fourier,noise reduction is performed,and then the threshold is extracted to be detected The pixels of the whistle signal are connected with the broken whistle using the eight-connectivity and Bresenham drawing method.Then by extracting the characteristics of the whistle as a data set,it is sent to the linear discriminator and the K nearest neighbor classifier for classification.This method provides a detection and recognition method for extracting the complete whistle spectrum contour and understanding the whistle signal of unknown features.4.This paper studies a method to recognize the calls of cetaceans in various kinds of ocean sounds.There are many kinds of sound in the ocean.Five kinds of sound that can represent the main sound in the ocean are selected.The audio is made into time-frequency map samples.A data amplification strategy for ocean sound is designed.The original data samples are amplified and convolutional neural network is built.The influence of different activation functions on the performance of the final convolutional neural network recognition model is compared.In conclusion,this paper comprehensively studies the cetacean call detection and recognition methods,and through the simulation experiment proves the feasibility of each algorithm,which lays a foundation for understanding the biological behavior of cetaceans and protecting cetaceans.