阅读说明：本技术 频率响应方法和装置 (Frequency response method and apparatus ) 是由 安德烈亚斯·奥利弗·特尔 于 2018-12-12 设计创作，主要内容包括：本发明提供了一种用于对时间信号进行滤波的方法和装置。根据K个权重l<Sub>k</Sub>的列向量l来指定(101,201)频率f的目标幅度频率响应H<Sub>T</Sub>(f),其中log H<Sub>T</Sub>(f)＝l<Sup>T</Sup>W(f)并且W(f)是K个幅度基函数W<Sub>k</Sub>(f)的列向量。计算(102,214)由log H<Sub>C</Sub>(f)＝g<Sup>T</Sup> V(f)定义的约束频率响应H<Sub>C</Sub>(f),其中V(f)是N个约束基函数V<Sub>n</Sub>(f)的列向量,对于V<Sub>n</Sub>(f),每个exp g<Sub>n</Sub>V<Sub>n</Sub>(f)满足通过级联保留的约束,并且g是满足l<Sup>T</Sup>W(f)和g<Sup>T</Sup> V(f)之间的匹配标准的N个系数的列向量。接收(103,212)输入时间信号,并用约束频率响应H<Sub>C</Sub>(f)对其滤波(104,210),以形成滤波的时间信号；以及输出(105,211)滤波的时间信号。(The invention provides a method and a device for filtering a time signal. According to K weights l k To specify a target amplitude frequency response H of the (101, 201) frequency f T (f) Wherein log H T (f)＝l T W (f) and W (f) are K amplitude basis functions W k (f) The column vector of (2). Computing (102, 214) from log H C (f)＝g T V (f) defined constrained frequency response H C (f) Where V (f) is N constraint basis functions V n (f) For V, for the column vector of n (f) Each exp g n V n (f) Satisfy pass levelConstraint of union reservation, and g is satisfied with l T W (f) and g T V (f) a column vector of N coefficients of the matching criterion. Receiving (103, 212) an input time signal and using a constrained frequency response H C (f) Filtering (104, 210) it to form a filtered time signal; and outputting (105, 211) the filtered time signal.)

1. A method of filtering a time signal, the method comprising implementing on a digital signal processing device the steps of:

according to K weights l_kColumn vector l to specify a target amplitude frequency response H of frequency f_T(f) Wherein log H_T(f)＝l^TW (f), and W (f) are K amplitude basis functions W_k(f) A column vector of (a);

calculation by log H_C(f)＝g^TV (f) defined constrained frequency response H_C(f) Wherein V (f) is N constraint basis functions V_n(f) For V, for the column vector of_n(f) Each exp g_nV_n(f) Satisfy the constraint of reservation by concatenation, and g is satisfied with l^TW (f) and g^TV (f) a column vector of N coefficients of a matching criterion;

receiving an input time signal;

with said constrained frequency response H_C(f) Filtering the input time signal to form a filtered time signal; and

outputting the filtered time signal.

2. The method of claim 1, wherein the matching criteria is with respect to/^TW (f) and g^TV (f) one or more error metrics between.

3. The method of claim 2, wherein the one or more error metrics comprise,/, measured by^TW(f)-g^TRe V (f) defined measure e (f).

4. The method of claim 3, wherein the condition of the matching criterion is a minimization of variation with respect to g, and the metric e (f) is an inner product<e,e>Wherein, in the step (A),<a,b>is the integration or summation over f between functions a (f) and b (f), whereby g satisfies the matching criterion as Ml, wherein,M＝η⁺c, η is composed of<Re V_n,Re V_p>Defined N × N matrix η_npThe superscript + denotes an inverse or pseudo-inverse operation, and C is represented by<Re V_n,W_k>Defined N × K matrix C_nk。

5. The method of claim 4, wherein the inverse or pseudo-inverse operation is a Moore-Penrose pseudo-inverse.

6. The method according to any one of claims 1 to 3, wherein g is specified as g ═ Ml, where M is a matrix of N × K, and the matching criteria include the condition that M is calculated to satisfy one or more error metrics with respect to an error function vector E (f) defined by E (f) ═ W (f) -M Re V (f).

7. The method of claim 6, wherein the inner product represented by < a, b > is an integral or sum over f between functions a (f) and b (f), and the one or more metrics incorporate E (f) into the inner product as at least one parameter <, E >.

8. The method of claim 7, wherein the one or more error metrics are for at least some k, j combination<E_k,E_j>。

9. The method of claim 8, wherein the one or more error metrics comprise for each k<E_k,E_k>And the condition is a minimization of a variation with respect to M.

10. The method of claim 7, wherein the one or more metrics are for each k<W_k,E_k>And the condition is<W_k,E_k>0, whereby when M ═ C⁺θ time M satisfies the matching criterion, where matrix C of K × N is C_k,n＝<W_k,ReV_n>The superscript + denotes an inverse or pseudo-inverse operation, and the matrix θ of K × K is θ_k,j＝<W_k,W_j>。

11. The method of claim 10, wherein the inverse or pseudo-inverse operation is a Moore-Penrose pseudo-inverse.

12. The method of any one of claims 1 to 11, wherein:

the step of specifying the target amplitude frequency response further comprises dynamically changing the target amplitude frequency response according to user preferences or operations of a computational auditory perception model, an

The step of calculating the constrained frequency response further comprises dynamically updating the calculation of the constrained frequency response after a dynamic change in the target amplitude frequency response.

13. The method of any of claims 1 to 12, wherein the constraint that is preserved by concatenation is a minimum phase constraint.

14. The method according to any of claims 4, 5 or 7 to 11, wherein the integrand of the inner product integration or summation is multiplied by a weighting function λ (f) adapted to the desired frequency enhancement.

15. The method of any one of claims 1 to 14, wherein V_n(f) Comprises a derivative of the gain with respect to the logarithm of the recursively implemented frequency response of the band filter.

16. The method of any of claims 1-15, wherein the steps of specifying the target amplitude frequency response and calculating the constrained frequency response are repeated as needed.

17. The method of any one of claims 1 to 16, wherein the time signal is an audio signal or a digital communication signal.

18. The method of claim 17, wherein the time signal is an audio signal.

19. An apparatus for filtering a time signal, the apparatus comprising a digital signal processing device comprising:

a target amplitude frequency response designator configured to designate according to K weights l_kColumn vector l to specify a target amplitude frequency response H of frequency f_T(f) Wherein log H_T(f)＝l^TW (f), and W (f) are K amplitude basis functions W_k(f) A column vector of (a);

a constrained frequency response determiner programmed to compute a log H_C(f)＝g^TV (f) defined constrained frequency response H_C(f) Wherein V (f) is N constraint basis functions V_n(f) For V, for the column vector of_n(f) Each exp g_nV_n(f) Satisfy the constraint of reservation by concatenation, and g is satisfied with l^TW (f) and g^TV (f) a column vector of N coefficients of a matching criterion;

an input time signal receiver interface for receiving an input time signal;

a filter for using the constrained frequency response H_C(f) Filtering the input time signal to form a filtered time signal; and

a filtered time signal output interface for outputting the filtered time signal.