CN108848044A - A kind of extracting method of channel fine feature - Google Patents

Abstract

The invention belongs to the technical field of signal sorting, and in particular relates to a method for extracting subtle features of a channel. The transmitted signal is affected by the channel during the transmission process, and the received signal implies subtle features that can represent channel information. In the present invention, the blind equalization algorithm is used to restore the transmitted signal from the received signal, and the peak value of the blind equalization weight vector is extracted, which is used as the subtle feature of the channel for the description and distinction of the channel. Through simulation experiments, Embodiment 1 verifies the feasibility of the method for extracting subtle channel features.

Description

A Method for Extracting Subtle Features of Channel

technical field

The invention belongs to the technical field of signal sorting, and in particular relates to a method for extracting subtle features of a channel.

Background technique

In mobile communications, the electromagnetic path through which signals propagate is called a radio channel. Wireless channels are closely related to the surrounding environment, and wireless channels in different environments have some differentiated characteristics. How to discover and extract signal-related features in wireless channels has great research value and is also a current research hotspot. Common wireless channel characteristic parameters include multipath delay, Doppler frequency shift, etc. At present, there are three main types of feature extraction methods for wireless channels: spectral estimation, estimation methods based on parameter subspaces, and deterministic parameter estimation.

For spectrum estimation, the common method is Multiple Signal Classification (MUSIC). It provides asymptotically unbiased estimates of the number of incident wavefronts, the direction of arrival or emission, the strength and cross-correlation of the incident waveform. However, this algorithm requires a large amount of calculation and storage for parameter space search, and when the incident signal is a coherent signal, the MUSIC algorithm is invalid.

The parameter subspace estimation method, the representative of this type of algorithm is the Estimating Signal Parameters via Rotational Invariance Techniques (ESPRIT) algorithm. It exploits the rotation-invariant property of the signal subspace and can be used for accurate parameter extraction of the horizontal angle. ESPRIT avoids the space search process and requires less calculation, but requires that the corresponding array elements of the subarray must be exactly the same dipole pair.

Deterministic parameter estimation, such as Expectation Maximization (EM), which is generated based on Maximum Likelihood Estimation (MLE). It can realize the joint estimation of time delay, horizontal angle and amplitude. The extension of EM algorithm is Space-Alternating Generalized Expectation Maximization (SAGE). The SAGE algorithm can jointly estimate multiple parameters such as time delay, angle of departure, angle of arrival, Doppler frequency shift, and amplitude at the same time.

The channel is the necessary link between the transmitting end and the receiving end. During the transmission process, the transmitted signal is affected by the channel background environment, and the received signal at the receiving end contains channel information. When there is a difference in the channel background environment, this difference is also reflected in the received signal. Therefore, subtle features that can reflect channel information can be extracted from the received signal. Let x(t) be the transmitted signal, and h(t) be the equivalent baseband impulse response, which includes the total transmission characteristics of the radio frequency and intermediate frequency parts of the transmitter, channel and receiver, then the received signal can be expressed as:

in, Represents convolution.

By adopting the idea of recovering x(t) from y(t) through equilibrium, the information of h(t) can be obtained, and the direct calculation of h(t) can be avoided. If the impulse response of the equalizer is ω(t), then the output of the equalizer is:

In the formula, g(t) is the total equivalent impulse response of the transmitter, the channel, the radio frequency of the receiver, the intermediate frequency part and the equalizer.

The ideal output value (that is, the expected output value) of the equalizer is the transmitted signal x(t), and in order to make z(t)=x(t), it must satisfy:

The purpose of the equalizer is to realize the above formula, and its frequency domain expression is:

H(f)·Ω(f)=1

The above formula shows that the equalizer is actually an inverse filter of the transmission channel. Therefore, consider the channel information represented by the equalizer weight vector.

Contents of the invention

Aiming at the problem of channel feature extraction, the present invention provides a method for extracting subtle channel features, which can extract channel information from received signals. Compared with the existing channel feature extraction methods, this method has a small amount of calculation and is easy to implement.

In order to achieve the above-mentioned purpose of the invention, the extraction process of channel fine features of the present invention includes the following steps:

S1: Let the received signal be N is the data length, which is normalized, μ is the mean value of the received signal, σ is the standard deviation of the received signal, and the normalization formula is:

S2: Set the initial value of the iteration step size μ and the blind equalization weight vector f _i (k).

S3: Use the improved MCMA algorithm to perform blind equalization on the data to generate an error signal e(k), whose formula is:

e(k)=e _R (k)+je _I (k)

in,

S4: The number of iterations of the blind equalization weight vector is k=N, and the iterative update formula is:

f _i (k+1)=f _i (k)+μe(k)x(ki)

Among them, e(k) is the error signal, μ is the iteration step size, i is the number of taps, and k is the number of iterations.

S5: The blind equalization tap coefficient residual mean square error rms _k is used as the evaluation index of equalization performance, and its formula is:

Among them, f _k,i is the ith element of the blind equalization weight vector f _i (k).

S6: Extract local maxima for the blind equalization weight vector f _i (k) modulus, and form all local maxima into a peak set

Where ||F|| represents the number of elements contained in the set F. is the sequential peak point, satisfying: Define a sequence of adjacent distances for sequential peak points:

S7: Based on the peak set F, extract the following subtle features:

Peak-to-average ratio: F _a =max(F)/F _harmmean , where

Peak-to-peak ratio: F _b =max(F)/median(F)

Peak point distribution uniformity: Where ||F|| represents the number of peaks.

The beneficial effect of the present invention is that the method for extracting channel subtle features of the present invention utilizes the reflection of channel background environment differences in received signals; in fact, in the present invention, the transmitted signals pass through different channels, resulting in differences in received signals. Therefore, the subtle features that characterize channel information can be used to describe and distinguish channels.

Description of drawings

Fig. 1 is a simulation block diagram of the present invention;

Fig. 2 is the flowchart of the realization process of the present invention;

Fig. 3 is a blind equalization weight vector of different received signals in Embodiment 1 of the present invention.

Detailed ways

Specific embodiments of the present invention will be described below in conjunction with the accompanying drawings, so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when detailed descriptions of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

The feasibility of the channel subtle feature extraction method is verified by the simulation experiment data, and the simulation block diagram is shown in Figure 1.

The experimental hardware platform includes a desktop computer with Pentium(R) Dual-Core 3.2G processor and 6G memory. The software platform is WIN7 operating system and Matlab2015b.

Fig. 2 is the implementation steps of the present invention in application and channel fine feature extraction. As shown in Figure 2, the method for extracting subtle channel features includes the following steps:

S1: Let the received signal be N is the data length, which is normalized, μ is the mean value of the received signal, σ is the standard deviation of the received signal, and the normalization formula is:

S2: Set the iteration step size μ=0.001 and the initial value of the blind equalization weight vector f _i (k) to be 1 for the center tap, and to be 0 for the rest. S3: Use the improved MCMA algorithm to perform blind equalization on the data to generate an error signal e(k), whose formula is:

e(k)=e _R (k)+je _I (k)

in,

S4: The number of iterations of the blind equalization weight vector is k=N, and the iterative update formula is:

f _i (k+1)=f _i (k)+μe(k)x(ki)

Among them, e(k) is the error signal, μ is the iteration step size, i is the number of taps, and k is the number of iterations.

S5: The blind equalization tap coefficient residual mean square error rms _k is used as the evaluation index of equalization performance, and its formula is:

Among them, f _k,i is the ith element of the blind equalization weight vector f _i (k).

S6: Extract local maxima for the blind equalization weight vector f _i (k) modulus, and form all local maxima into a peak set

Where ||F|| represents the number of elements contained in the set F. is the sequential peak point, satisfying: Define a sequence of adjacent distances for sequential peak points:

S7: Based on the peak set F, extract the following subtle features:

Peak-to-average ratio: F _a =max(F)/F _harmmean , where

Peak-to-peak ratio: F _b =max(F)/median(F)

Peak point distribution uniformity: Where ||F|| represents the number of peaks.

Example 1

In this embodiment, the feasibility of the channel subtle feature extraction method is verified by using simulation experiment data. The sending signal adopts QPSK modulation mode, and the data length is 100,000 points. In the simulation, the channel is a simulated Rayleigh channel, and the orders of h1, h2, and h3 are different. The simulation parameter settings are shown in Table 1:

Simulation parameter setting situation in the embodiment 1 of table 1

As shown in Figure 3. h1, h2, and h3 respectively correspond to the modulus of the last weight vector obtained by the 100,000-point normalized received data of the three channels in Table 1 after the iterative processing of the improved MCMA algorithm blind equalization, and h1, h2, and h3 are all 44-order equalizers Balanced results. It can be seen from Fig. 3 that there are differences in blind equalization results of signals received in different channels.

As shown in table 2:

Subtle feature extraction results of different channels in Table 2 Example 1

subtle features F _a F _b F _p h1 11.6350 7.6917 16.7225 h2 3.7839 2.5320 19.0195 h3 4.8850 3.5450 17.8421

Comparing the three channels h1, h2, and h3, there are differences in the values of the three subtle features of peak-to-average ratio F _a , peak-to-peak ratio F _b , and peak point distribution uniformity F _p . Therefore, the channel is different, and the subtle characteristics of the channel will also be different. Therefore, subtle features: peak-to-average ratio F _a , peak-to-peak ratio F _b , peak point distribution uniformity F _p , can be used to describe and distinguish different channels.

Claims (1)

1. A method for extracting channel fine features is characterized by comprising the following steps:

s1, setting the received signal asN is the data length, normalized:

wherein, mu is the mean value of the received signal, and sigma is the standard deviation of the received signal;

s2, setting iteration step size mu and blind equalization weight vector f_i(k) An initial value of (1);

s3, carrying out blind equalization on the data by adopting an improved MCMA algorithm to generate an error signal e (k):

e(k)＝e_R(k)+je_I(k)

wherein,

s4, taking k as N as the iteration number of the blind equalization weight vector, and the iteration update formula is:

f_i(k+1)＝f_i(k)+μe(k)x(k-i)

wherein e (k) is an error signal, i is the number of taps, and k is the number of iterations;

s5: residual mean square error rms using blind equalized tap coefficients_kAs evaluation indexes of balance performance:

wherein f is_k,iFor blind equalization of weight vectors f_i(k) The ith element of (1);

s6: for blind equalization weight vector f_i(k) Extracting local maximum values by the module values, and forming a peak value set by all the local maximum values:

where F represents the number of elements included in set F,for the sequential peak points, satisfy:defining a sequence of adjacent distances of sequential peak points:

s7, extracting the following fine features based on the peak set F:

peak-to-average ratio: f_a＝max(F)/F_harmmeanWherein

Peak to peak ratio: f_b＝max(F)/median(F)；

Uniformity of peak point distribution:wherein F represents the number of peaks.