Design Of Lattice Network Based Digital Filters With High Robustness

With the development of computer and electronic technology, digital signal processing(DSP) technology develops quickly and has been extensively used in many fields. In digitalsignal processing, digital filters are playing an important role for their superiority over analogfilters, such as high precision, high reliability and easy large-scale integration. Structure designof digital filters is an important subject in DSP area.It is well known that a filter can be implemented with many di?erent structures. In?u-enced by finite word length e?ect, filters implemented with di?erent structures may yield verydi?erent figures of performance. So, design of digital filters with high robustness and lowcomplexity has been an important topic in recent decades. In this thesis, a novel lattice filterstructure is derived by combining the taped numerator and injected numerator lattices. Oneremarkable feature of the proposed structure is that, for an Nth order filter, the input signalcan be injected at 2N nodes arbitrarily, which allows us to optimize the structure with respectto these injector points. The two types of the one-multiplier lattices are chosen arbitrarily ateach stage of the new structure, it can not only reduce the implementation complexity, but alsooptimize the new structure further.Simulations are carried out on MATLAB plat-form to compare the performance betweenour new structure and those traditional structures on signal power ratio and parameter sensi-tivity. It is shown that the proposed structure is not only canonic in the number of multipliersbut also possesses much improved finite word length properties such as very low parametersensitivity and very uniform signal power ratio across signal nodes.