A method and device for an all-phase filter bank configured using frequency response characteristics

Abstract

The invention discloses a method and device for configuring a full-phase filter bank through the frequency response characteristic. The method comprises the steps of convolving an input signal x(n) through a convolution window, adding every two multiplied data with the interval of N-1 to form N data, rapidly configuring the frequency response characteristic to Q subchannel filter coefficient vectors, obtaining the configured Q subchannel filter coefficient vectors, and carrying out multiply-accumulate on the data to obtain Q filtering output results. The device is characterized in that signals to be filtered are sampled through an analog-to-digital converter to obtain a sample sequence, the sample sequence enters a DSP device in a parallel data input mode, the filter order N, the number p of 0 and the number m of 1 in a frequency vector H are set, and Q filtering output results are obtained through processing of the DSP device. According to the method and device for configuring the full-phase filter bank through the frequency response characteristic, rapid configuration of the locations of sub-bands and the widths of the sub-bands of a filter, a filter coefficient and a frequency response function is achieved, and signal characteristics can be accurately extracted.

Description

technical field [0001] The present invention relates to the technical field of digital signal processing, in particular to a method and device for quickly and effectively building a corresponding all-phase filter bank according to provided frequency response characteristics. Background technique [0002] in voice [1] ,Troubleshooting [2] ,radar [3] , sonar, earthquake, imagery [4] , communication, control, biomedicine, remote sensing and telemetry [5] , geological exploration, aerospace, automation instrumentation [6] In other fields, there are a large number of signal feature extraction problems. A commonly used perspective of these feature extraction problems is to analyze the relative energy in each frequency band contained in the signal. For example, in speech signal recognition [7][8] Among them, the frequency band of human voice is distributed in the range of [300Hz, 3400Hz], and the voice difference of different people is reflected in the difference in energy d...

AI Technical Summary

Problems solved by technology

However, the way of dividing the frequency bands of the wavelet filter bank method is fixed, that is, the average sub-band method is adopted from the upper level decomposition to the lower level decomposition, and its parameter configuration is not flexible; while multi-level decomposition means multiple filtering operation, a large amount of calculation

Images