Method for calculating coefficients of filter and method for filtering
AssigneeSunplus mMobile Inc.
InventorsWei-De Wu, Shin-Yuan Wang
Life of this patent
Estimating the term…
Abstract
A method for calculating coefficients of a filter and a method for filtering are provided. The invention directly factorizes a specific function in a cepstrum domain by spectral factorization and cepstrum technique to obtain coefficients of denominator function from the filter. In other words, the invention adopts a non-iterative algorithm to reduce computational complexity and avoid convergence due to calculating coefficients. Besides, the specific function of the invention includes a compensation function, so that a Fourier transform with greatly reduced size can be utilized in the spectral factorization to greatly save the computations and keep good system performance at a receiver.
Claims14
A method for calculating coefficients of a filter, wherein the filter is adapted to an equalizer of a receiver to eliminate interferences to a received signal from the receiver caused by transmission channel, the method comprising:
- a. estimating a channel response;
- b. estimating a signal-to-interference-plus-noise power ratio (SINR);
- c. calculating an autocorrelation function of the channel response;
- d. summing up the autocorrelation function of the channel response, a reciprocal of the SINR, and a compensation function, so as to obtain a specific function;
- e. retrieving a plurality of coefficients from the specific function;
- f. performing a cepstrum computation on the plurality of coefficients, so as to correspondingly obtain a plurality of quefrency coefficients;
- g. selecting a part of the plurality of quefrency coefficients to perform a Fourier transform with a limited size, so as to obtain a plurality of spectral components;
- h. applying an exponential computation on the spectral components, so as to obtain a plurality of exponential spectral components;
- i. applying an inverse Fourier transform with the limited size on the exponential spectral components, so as to obtain a plurality of coefficient candidates;
- j. selecting a part of the plurality of coefficient candidates, and normalizing the plurality of coefficient candidates to serve as coefficients of a denominator function of the filter, so as to form the denominator function;
- and
- k. calculating coefficients of a numerator function of the filter according to the coefficients of the denominator function and the channel response.
A method for filtering, comprising:
- receiving a signal from transmission channel:
- calculating a denominator function and a numerator function of filter, wherein a length of the denominator function is expressed as L, the coefficients of the denominator function are expressed as {an}n=0L−1;
- filtering the signal according to the numerator function by using a feed-forward filter, so as to obtain a numerator filtered signal, wherein the numerator filtered signal is expressed as {vn}n=−LL−1;
- providing a threshold value, wherein the threshold value is expressed as η/L, wherein η is a preset positive integer;
- sequentially determining whether a plurality of coefficients in the denominator function is smaller than the threshold value;
- selecting and setting the coefficients smaller than the threshold value to zero from the plurality of coefficients, and further adjusting the denominator function into an equivalent denominator function, wherein the equivalent denominator function is
- a
- ⋓
- n
- =
- {
- a
- n
- ,
- if
-
-
-
- a
- n
-
- 2
- ≥
- η
- /
- L
- 0
- ,
- if
-
-
-
- a
- n
-
- 2
- <
- η
- /
- L
- ;
- and
- filtering the numerator filtered signal according to the equivalent denominator function by using a feedback filter, so as to obtain a final filtered signal, wherein the final filtered signal is expressed as
- x
- ^
- n
- =
- {
-
- v
- n
- -
- ∑
- m
- =
- 1
- L
- -
- 1
-
- a
- ⋓
- m
-
- x
- ^
- n
- -
- m
- ,
- if
-
-
- n
- >
- 0
-
- 0
- ,
- if
-
-
- n
- ≤
- 0.
Dependencies are inferred from the claim text.