Research On FFT Processing Algorithm Based On Hybrid Filter Bands

How to realize low-complexity and high-precision large-point FFT algorithm becomes the key of signal processing system with limited resources.The method based on filter banks has better performance of low implementation complexity and high computational accuracy.In this paper,two FFT processing algorithms based on hybrid filter banks are proposed to achieve high precision and low complexity FFT under different requirements.Chapter 2 studies the basic theory of FFT algorithm and hybrid filter bank.The principle and implementation process of the two-dimensional FFT algorithm are analyzed,and it is pointed out that the disadvantage of the algorithm is that it does not increase the operation rate.At the same time,the shortcomings of the current optimal prototype filter method are analyzed,and the hybrid filter bank obtained by this method is introduced,which has the problem of low precision in FFT operation.Finally,two research ideas that can achieve low complexity and high precision are proposed targeting at the above two short coming.Chapter 3 proposes a low complexity FFT algorithm based on analysis filter bank.The algorithm utilizes the advantages of efficient multi-phase filter bank structure and half-band filter,which can reduce the number of points per FFT operation and reduce the operating frequency by downsampling.Simulation results show that,within the allowable range of the error,we can adjust the number and order of the decomposition filter banks,and the energy consumption can be reduced by 50%compared with the direct FFT algorithm and the two-dimensional FFT algorithm.Chapter 4 proposes a high precision FFT processing algorithm based on cosine-modulated filter banks.The algorithm designs the prototype filter into the form of U(zd)·V(z).By optimizing the cutoff frequency ?c of the prototype filter U(z),the window function method is used to indirectly design a prototype with high reconfigurability.The filter reduces the design difficulty and time of the high-order prototype filter.we can effectively suppress the amplitude distortion and aliasing distortion of the approximate perfect reconstructed sub-band filter bank obtained by cosine modulation,thereby improving the precision of the FFT processing.The simulation shows that under the premise of relaxing the energy-saving effect,the error value of the algorithm and the original spectrum of the signal can be controlled within-100dB,and the high precision of the operation is realized.