Fluctuation Characteristics Analysis Of Wind Speed Signal Based On The Sparse Representation Theory

Researching intelligent robots which can locate toxic gas leak sources contains a huge application value both in civil and military field.However,the complex and changeable actual airflow field environment brings great difficulties and challenges to the work of locating toxic gas sources by intelligent robots.Intensive study of the flow and evolution of complex gaseous flow field will provide valuable clues for solving the above problems.In this thesis,the theory of sparse representation is mainly used to study the wind speed signal fluctuation characteristics of natural wind field and the fluctuation of wind speed signal in different environments.From the results we get some regular understanding of natural wind field.The specific works are as follows:Firstly,we test and analyze the performance of 5 algorithms in sparse representation theory.The test results of Basis Pursuit algorithm,Matching Pursuit algorithm and Orthogonal Matching Pursuit(OMP)algorithm show that the coefficients obtained by OMP algorithm are more sparse than others when the dictionary and error margin are same.The results of DCT analysis dictionary and K-SVD learning dictionary show that K-SVD learning dictionary has more powerful ability of characterizing because K-SVD dictionary can capture the internal characteristic of signal effectively by training samples.Secondly,because K-SVD dictionary can capture the detail fluctuation characteristics of the wind speed signal by continuously training wind speed signal samples.Therefore,this algorithm is used to extract the local detail characteristics from the indoor wind speed signal and the outdoor wind speed signal.At last,we analyze these local detail characteristics by the recurrence plot and recurrence quantification algorithm.The results show that the indoor wind speed signal and the outdoor wind speed signal have similar local detail characteristics.But in general,the indoor wind speed signal is more complicated than the outdoor wind speed signal.Thirdly,K-SVD dictionary ignores the correlation and continuity between adjacent signal segments when training,which results in the redundancy between dictionary atoms.The convolutional sparse coding takes full account of the local correlation of the signal,and it trains the entire wind speed signal directly.In each iteration,the filter searches for the fluctuation characteristics of the signal.Therefore,the convolutional sparse coding algorithm is used to obtain the filter dictionary of the wind speed signal by training the entire indoor wind speed signal and outdoor wind speed signal.These filters are viewed as the global fluctuation characteristics of the wind speed signal.Finally,these characteristics are analyzed by the recurrence plot and recurrence quantification.The results show that the indoor wind speed signal and the outdoor wind speed signal contain similar fluctuation components.But on the whole,the indoor wind speed signal is more complicated and more random than the outdoor wind speed signal.