Application Of Artificial Intelligence In The Study Of Marine Mammals' Acoustic Signals Detection And Classification

Driven by the need of bionic application and ecological civilization construction,marine mammal acoustic study has developed rapidly in these years.Currently,acoustic monitoring systems are widely used to obtain the data to conduct research,but the large scale of acoustic data makes the traditional manual processing method difficult to sustain.Detecting and classifying the marine mammals' acoustic signals from the vast amounts of original acoustic data have become the major barrier of marine mammal acoustic study.With the development of computer technology,artificial intelligence has played an important role in many research fields.However,the application of artificial intelligence in the study of marine mammals' acoustic signals is just beginning.In this paper,based on the measured marine mammal acoustic signal data,a method based on image processing is first proposed to detect odontocetes echolocation clicks,then a method based on convolutional neural network is proposed to identify marine mammal calls.Finally,the identification method is transplanted to RK1808 chip embedded platform,and the identification can be done in real-time.The main work and innovations are as followed:1)A method based on image processing is first proposed to detect odontocetes echolocation clicks.Experiments were carried out using the measured acoustic signal data of four animals.The results show that compared with the traditional energy operator method,the proposed method has higher recall rate and higher accuracy under the condition of low SNR and can achieve better detection effect.2)A method based on convolutional neural networks is proposed to identify marine mammal calls.In this method,the detection and classification are integrated in one step.Experimental results on the measured acoustic signal data of three kinds of animals show that it has high recognition recall rate and recognition accuracy.3)Aiming at the problems of high cost and high power consumption of the existing deep learning embedded platform,the hardware design and construction of the embedded platform are completed by using the domestic RK1808 chip.The recognition algorithm proposed in 2)is transplanted to the platform to realize the real-time recognition of animal acoustic signals and achieve the ideal recognition speed and accuracy.4)Developed an intelligent identification system for marine acoustic data with graphical interface,and packaged relevant algorithms in the form of application software,so that users can easily and quickly use the system to identify a variety of Marine mammal calls,ship noises and windy weather.After testing,the system can stably complete the established functions...