The non-stationary signal analysis is one of the significant and difficult domains in signal processing technology. Its aim is to analyze and process the locality of non-stationary signals and describe the signals with some time-varying statistics. This dissertation summarizes the methods for analyzing non-stationary signals, introduces the two algorithms of adaptive signal decomposition, i.e. empirical mode decomposition and local narrow band signal decomposition, and describes the relationship between the two algorithms. It also proposes 2D empirical mode decomposition to process 2D texture images, and at the same time, modifies some defects of empirical mode decomposition. This dissertation also makes researches in the application domain of non-stationary signal analysis. First of all, it represents images as AM-FM combined with bidimensional empirical mode decomposition and image AM-FM model, and proposes methods for estimating the important parameters of the model. Secondly, instantaneous frequency is a significant measurement for non-stationary signals. The estimation of instantaneous frequency had been applied into many fields. This dissertation introduces the concept of instantaneous frequency and estimating technology of instantaneous frequency in detail, and then, proposes a new method, i.e. the direct method, to estimate instantaneous frequency based on empirical mode decomposition and local narrow band signal decomposition. Theory analysis and experiment results indicate that the algorithms proposed in this thesis have respective advantages, such as innovation, accuracy, simplicity, applicability, etc.
修改评论