Multirate Signal Processing And Its Applications To Real Time System

Multirate signal processing is an important area in modem signal processing theory. During the past decades, a great development has been achieved in this area, and many successful applications could be found in various fields. This thesis concentrates on the analysis of sampling rate changing (decimation and interpolation); the discussion of basic multirate signal processing theory; the implementation of high efficient algorithm, which provides the bases for other algorithm development in this thesis.Uncertainty Principle points out that , which means that high time resolution and high frequency resolution can not be obtained both at the same time in spectrum analysis. Integrating multirate signal processing theory and classical spectrum estimation theory, a brand new multiresolution spectrum theory is provided and a special algorithm is discussed in great detail. The polyphase decomposition theory is also applied to optimize the algorithm and improve the computing efficiency.The multiresolution method is extended and applied to random vibration control system, thus a multiresolution random vibration control algorithm is created, which provides a satisfied solution to the poor control precision problem at low frequency that was pointed out and remained unsettled for many years. The control algorithm has been successfully implemented on Dactron's shaker control system, showing great improvement in low frequency control precision, with a little more computation resource consumed.High resolution octave analysis is another successful application of multiresolution spectrum estimation theory. Octave analysis is widely used in acoustic signal analysis, whose resolution needs to be enhanced greatly in order to describe the sound signal in more detail. Today multi-channel real-time octave analysis may need to be processed at 1/48-octave or even 1/96-octave, which is impossible for some platforms unless more expensive hardware enhancement provided. The high-resolution octave analysis algorithm provided in this thesis is based on the multiresolution spectrum estimation theory. It is a well-compromised solution to both the computation resource problem and measure time problem. Compared to the integrate average method, which is found widely used in octave analysis, this algorithm saves much computation.DSP-based real time hardware design is also studied in this thesis. A hardware platform combining DSP, FPGA and USB is described, and several technical issues are discussed and then" practical solutions are provided. Together with the signal processing theory mentioned above, the hardware design method presents a complete solution to real time multirate signal processing.In a word, this thesis makes breakthroughs in both theory and application. In theory, multiresolution spectrum estimation method, multiresolution random vibration control algorithm and high resolution octave analysis algorithm are developed; in practice, two types of DSP hardware are designed and implemented with all the algorithms mentioned above, all of them meet product level demands.