US 11,816,180 B2
Method and apparatus for classifying mixed signals, and electronic device
Zhiyong Feng, Beijing (CN); Kezhong Zhang, Beijing (CN); Zhiqing Wei, Beijing (CN); Li Xu, Beijing (CN); and Che Ji, Beijing (CN)
Assigned to BEIJING UNIVERSITY OF POSTS AND TELECOMMUNICATIONS, Beijing (CN)
Appl. No. 17/602,692
Filed by Beijing University of Posts and Telecommunications, Beijing (CN)
PCT Filed Mar. 10, 2020, PCT No. PCT/CN2020/078573
§ 371(c)(1), (2) Date Oct. 12, 2021,
PCT Pub. No. WO2020/215911, PCT Pub. Date Oct. 29, 2020.
Claims priority of application No. 201910328208.9 (CN), filed on Apr. 23, 2019.
Prior Publication US 2022/0180095 A1, Jun. 9, 2022
Int. Cl. G06K 9/00 (2022.01); G06K 9/62 (2022.01); G06F 18/2135 (2023.01)
CPC G06F 18/2135 (2023.01) [G06F 2218/12 (2023.01)] 8 Claims
OG exemplary drawing
 
1. A method for classifying mixed signals, comprising:
receiving mixed signals containing noises and at least two different types of signals;
performing calculation on a matrix corresponding to the mixed signals by means of a preset Principal Component Analysis PCA method to obtain to-be-classified mixed signals and to determine the number of types of signals contained in the to-be-classified mixed signals; wherein the to-be-classified mixed signals are mixed signals obtained after removing the noises in the mixed signals;
determining a separation matrix based on the number of types of signals contained in the to-be-classified mixed signals;
separating individual signals in the to-be-classified mixed signals by means of the separation matrix to obtain to-be-identified signals;
calculating a preset number of high-order cumulants corresponding to each to-be-identified signal in the to-be-identified signals respectively;
taking the calculated high-order cumulants as characteristics of the to-be-identified signal corresponding to the high-order cumulants respectively;
inputting the characteristics of the to-be-identified signal into a preset classification model; wherein the classification model is used for calculating and outputting a modulation mode of the to-be-identified signal based on the characteristics of the to-be-identified signal; and
obtaining an output result of the classification model; wherein the output result comprises the modulation mode of the to-be-identified signal;
wherein performing calculation on the matrix corresponding to the mixed signals by means of the preset Principal Component Analysis PCA method to obtain to-be-classified mixed signals and to determine the number of types of signals contained in the to-be-classified mixed signals, comprises:
normalizing a matrix R corresponding to the mixed signals to calculate a matrix R; wherein the matrix R is a matrix obtained by normalizing the matrix R corresponding to the mixed signals;
performing centralization processing on the matrix R, so that an average value of the matrix R is 0, in order to calculate a matrix R;
calculating an autocorrelation matrix of the matrix R; performing singular value decomposition on the autocorrelation matrix custom character(R·R″) of the matrix R to obtain custom character(R·R″)=ÛΛÛ″, wherein R″ is a transposed conjugate matrix of the matrix R, Û″ is a transposed conjugate matrix of Û, Û=[û1, . . . , ûN] is an orthogonal matrix, and ûN is a n-th column of the matrix Û, a diagonal matrix Λ is

OG Complex Work Unit Math
N is the number of antennas receiving the mixed signals; and λ1, . . . , λN are singular values of the autocorrelation matrix custom character(R·R″;
arranging the singular values λ1, . . . , λN from small to large; and setting singular values, whose numerical values are less than a preset threshold, among the singular values λ1, . . . , λN to 0, and calculating a diagonal matrix

OG Complex Work Unit Math
letting Ũ=[û1, . . . , ûN], calculating the to-be-classified mixed signals using a preset formula R≙Ũ·[rl, . . . , rL]; the matrix corresponding to the to-be-classified mixed signals is R; and
determining the number of types of signals contained in the to-be-classified mixed signals based on the number of singular values that are not 0 among the singular values λ1, . . . , λN.