I’m a Software Engineer at T-Head Alibaba, a member of the Algorithm and Framework Group.
I earned my Bachelor’s degree in Computational Information Science from Dalian University of Technology, followed by a Ph.D. in Acoustics from the Institute of Acoustics at the Chinese Academy of Sciences.
My research interests are algorithms, Physics and GPGPU. I’m also interested in more applied problems with nice theoretical components.
Here is my CV.
Passive ocean acoustic thermometry (POAT) needs long accumulation time to achieve high accuracy. This article provides a machine learning-based method, Random Forest, to obtain the averaged sound speed (AVSS). With supervised learning, the AVSS can be estimated from half an hour accumulated noise cross-correlation functions (NCFs). Based on the feature importance analysis, an empirical equation is proposed to briefly describe the relationships between the features. The results of estimations are compared among different methods to demonstrate the advantage of the machine learning-based approach.
Estimation of current speed from ocean ambient noise
For the observation of short-time ocean current, a method for estimating the velocity of ocean current using ambient noise is proposed in shallow water based on the passive acoustic tomography. The energy accumulation of noise cross-correlation functions can be increased by beamforming. The empirical Green''s functions between two horizontal arrays are extracted from the noise cross-correlation functions, and the time arrival structures of the empirical Green''s functions are used to invert the current velocity between arrays. By processing the experimental data, the empirical Green''s functions and current velocities are extracted for 2 hours, and the velocity variation of ocean current can be observed. The simulation demonstrates the feasibility of the method in this experimental environment, and the mismatch of depth and range is analyzed. Its effect is negligible.
Shallow water passive acoustic tomography using dictionary learning
A dictionary learning-based method for sound speed profile (SSP) inversion in shallow water is presented. The empirical Green's functions between two parallel horizontal arrays can be extracted from noise cross-correlation functions. The sound speed profiles are sparsely characterized by data-generated dictionary matrix, and they can be inverted by searching for sparse coefficients. This method is validated by experimental data in the South China Sea. Compared with the traditional empirical orthogonal function (EOF) methods, the inversion accuracy accuracy is reduced to 0.53 m·s-1, moreover this method has fewer search parameters and higher accuracy.
