基于过采样的混合滤波器组性能优化

Performance Optimization of Hybrid Filter Bank Based on Oversampling

模拟误差是制约混合滤波器组信号重构精度的主要原因,如何降低由模拟误差导致的分解滤波器组系数误差成为了首要问题。引入过采样技术,研究在不同过采样率下8通道混合滤波器组的性能,寻找到过采样率最优值约等于7%。在不同模拟误差下对基于过采样的混合滤波器组的性能进行仿真,结果表明,原型结构和双阶型结构的混叠值相近,但后者对模拟误差的敏感度比前者大。信号重构阶段,在最小二乘法的基础上采用频带加权法进行误差校准。在1%模拟误差内,采用7%过采样率的原型混合滤波器组相比无过采样,平均混叠值下降了约50 dB,最大混叠值下降了约94 dB。仿真验证了引入过采样的有效性。

The resolution of signal perfect reconstruction of hybrid filter batik （HFB） is very sensitive to analog imperfections, how to reduce the analysis filter bank coefficients error caused by the analog imperfections is the primary problem. Oversampling technique is introduced into HFB, the performance of the 8 - channel HFB with different oversampling rates is studied, and simulation finds that the optimal value among the oversampling rates is about 7 %. The performance of HFB based on the oversampling is studied in terms of sensitivity to analog imperfections. The simulation results show that the original structure is a better candidate than the two-stage one. Synthesis stage uses frequency band weighting method based on the least squares method for error calibration. Compared with none-oversampling one, the performance of the original HFB with 7 % oversampling rate is improved, with 45 dB mean aliasing value reduction and 94 dB max aliasing value reduction. Simulation verifies that introducing the oversampling technique to HFB is effective.