Research Of Filterbanks With Integer Coefficients And Its Applications In Communications

Two-channel filterbanks with integer coefficients had already been widely researched during last decade. The design methods of linear phase, low-delay two-channel filterbanks with integer coefficients have some conclusions in papers. For the case of M-channel modulated filterbanks, however, only very few design methods have been reported.First of all, we introduce the fundamental theory of filterbanks, which including the polyphase realization and the PR condition of cosine-modulated filterbanks. Then, we start studying on integer modulated filterbanks, mainly on the implementation of M-band cosine modulation filterbanks with integer coefficients.In this paper, the Alfred Mertins' subspace approach [17] is used for reference on the implementation of integer coefficient filter banks. We improve the arithmetic basis on former design of filter prototype, and put forward an improved arithmetic, which can get better stopband attenuation, on the design of prototype filter. The new arithmetic uses the RQI for computing the best eigenvalue and eigenvector. When quantize the coefficients, we select the suitable weights coefficients which can guarantee to achieve the minimal objective function. The simulation results show that new prototype filter has better stopband attenuation, and the attenuation of first sideband improves 3 dB. Then we combine the integer prototype and modulation matrix to implement the integer modulation filter banks. The PR property is guaranteed throughout the whole design process. And the whole design process with integer arithmetic allows for fast and efficient computation on signal processors and in dedicated hardware.In the end, we introduce the application of integer coefficients filterbanks in multi-carrier communication.