CN101567730B - Signal estimation and detection method based on nonlinear transformation - Google Patents

Abstract

The invention relates to a signal estimation and detection method based on nonlinear transformation. Two assumptions are given: there is no wireless spectrum signal H ₀ : y(t)=n(t) and there is wireless spectrum signal H ₁ : y(t)=x(t)+n(t). Detecting whether there is a signal x(t) in the received signal y(t) includes the following steps: 1) nonlinear transformation, performing nonlinear transformation on the received broadband signal y(t) through a nonlinear converter, Obtain the nonlinear signal f(t); 2) Feature extraction, extract the relevant features of the signal from the nonlinearly transformed signal f(t), and provide it to the matched filter as prior knowledge; 3) Matched filter detection, matching The filter carries out matching detection to the received signal y(t) according to the prior knowledge obtained in step 2), and the matching result is used as the basis of judgment; 4) judgment, the decision device detects the matched filter according to the judgment threshold set As a result, a judgment is made to determine whether or not the signal x(t) exists. The advantages are: strong real-time performance, high accuracy of signal estimation and detection, simple system structure, low computational complexity, and short detection time.

Description

Signal Estimation and Detection Method Based on Nonlinear Transformation

technical field

The present invention relates to estimation and detection technology of random signal, more specifically relates to a signal estimation and detection method based on nonlinear transformation under low signal-to-noise ratio (SNR) environment.

Background technique

At present, the contradiction between the ever-increasing spectrum demand and the limited spectrum resources is becoming more and more prominent, which seriously restricts the development of wireless communication services. However, from the perspective of actual wireless spectrum operation, the allocated (authorized) wireless spectrum is idle to a considerable extent in time and space. According to the measurement data report on the wireless spectrum, the spectrum utilization rate of most wireless frequency bands is only 10%. %about. How to effectively solve the contradiction between the scarcity of spectrum resources and the low utilization rate of spectrum has become a key technology in wireless communication. UWB endowed with Cognitive Radio (CR) functions is recognized as an effective technical means for efficient use of wireless spectrum.

With its advantages of high speed, high performance, low power consumption and low cost, UWB has great development potential in high-speed short-distance wireless communication, especially streaming media business. Perceiving and detecting the surrounding wireless environment correctly is the premise of Cognitive Ultra-Wideband (CR-UWB) work. The current cognitive wireless signal perception and detection methods can be divided into three categories: energy detection, matched filter detection and cyclostationary feature detection. Energy detection is the most widely used spectrum detection method at present, which uses the interference temperature as the metric to find suitable spectrum holes. However, the interference temperature threshold of energy detection is difficult to determine, and when the signal is extremely weak, it is difficult to distinguish signal, noise and interference, which is not suitable for the application environment with low signal-to-noise ratio like UWB. The matched filter detection structure is simple and can achieve high signal detection accuracy, but it needs to know the prior knowledge of the detected signal, which limits the application of this detection method and is not suitable for use in the UWB environment. Based on the signal cyclostationary feature detection, the signal can be distinguished according to the peak position of the signal feature in the three-dimensional space of the spectrum. It does not need to know the prior knowledge of the detection signal, and it still has good detection performance in the case of very low SNR. However, based on cyclostationary feature detection, it is necessary to perform high-order (second-order and above) statistics on the signal, the statistical space of the signal must be complete, the calculation complexity is high, and the detection time is long. It is not suitable for some network real-time services such as streaming media services. In the UWB application environment with high performance requirements.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the above-mentioned prior art, solve the problem of lack of prior knowledge of the detected signal in the matched filter detection method, and design a signal perception that can be quickly, accurately and effectively implemented in a low signal-to-noise ratio environment Detection method.

Above-mentioned purpose is achieved through following technical scheme:

Two assumptions are made on the wireless spectrum signals existing in the surrounding environment: one is that there is no wireless spectrum signal H ₀ ; the other is that there is a wireless spectrum signal H ₁

H ₀ : y(t)=n(t)

H ₁ : y(t)=x(t)+n(t)

n(t) represents additive Gaussian noise, x(t) represents a licensed wireless spectrum signal, and the received signal is y(t), where 0≤t≤T. Estimate and detect whether there is a signal x(t) in the signal y(t), including the following steps:

1) nonlinear transformation, the received broadband signal y(t) is nonlinearly transformed by a nonlinear converter to obtain a nonlinear signal f(t);

2) Feature extraction, extracting relevant features of the signal from the non-linearly transformed signal f(t), and providing it to the matched filter as prior knowledge;

3) matched filter detection, the matched filter performs matching detection to the received signal y(t) according to the prior knowledge obtained in step 2), and the matching result is used as the basis for judgment;

4) Judgment, the decision unit judges the detection result of the matched filter according to the set judgment threshold, and judges whether the signal x(t) exists.

A further design is that the nonlinear converter in the step 1) is

f(t)＝ay ² (t)-by(t-τ ₁ )y(t+τ ₂ )-cy(t-τ ₃ )

Where a, b and c are weighting coefficients, and τ ₁ , τ ₂ and τ ₃ are delay correlation times respectively.

A further design is that the acquisition of the prior knowledge in the step 2) is:

1) The nonlinearly transformed signal f(t) is time-frequency transformed by a fast Fourier transformer to obtain a time-frequency transformed signal F(f)

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; j</span>}\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;ft</span>}}\mathrm{<span\; class="notranslate">\; dt</span>}<span\; class="notranslate">\; ;</span>$

2) Take the modulus (absolute value) of F(f) after the time-frequency transformation, and obtain the frequency f ₀ and 3dB bandwidth B corresponding to the maximum value accordingly, as the priori spectrum characteristics of the signal x(t), That is prior knowledge.

A further design is that the matching detection in the step 3) is to perform matched filtering on the received signal y(t) through a matched filter according to the prior knowledge, and output the filtered signal s(t)

$\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; d\&tau;</span>}$

The matched filter function h(t) is determined by the prior knowledge f ₀ , that is, h(t)=cos[2πf ₀ (Tt)]+sin[2πf ₀ (Tt)], where τ is the delay time.

Further design is, described step 4) decision threshold is set to η

η=α|F(f)| ² /B

The coefficient α will be set according to the false alarm rate and missing alarm rate required by the system.

The decision device detects the output filter signal s(t) according to the above-mentioned decision threshold η, if H ₀ is established, the received signal y(t)=n(t), which does not contain the signal x(t), can use y( t) corresponds to the frequency band for communication; if H ₁ is established, the frequency band corresponding to y(t) contains signal x(t), so the frequency band corresponding to y(t) cannot be used for communication.

The method of the invention applies the nonlinear transformation to the matched filter detection, estimates the detected signal through the nonlinear transformation, quickly extracts the prior knowledge of the signal, and realizes fast and accurate perception and detection of the signal. Specifically, it is the combination of nonlinear transformation and matched filter, the nonlinear transformation quickly extracts signal features, and the matched filter accurately perceives and detects the signal, which solves the problem of lack of prior knowledge of the detected signal in the matched filter detection method. Thereby produce following beneficial effect:

(1) Through nonlinear transformation, the rapid extraction of signal prior knowledge is realized, and the real-time performance is strong;

(2) Accurate signal detection is realized through matched filter detection;

(3) The judgment threshold has an adaptive function, which can adapt to the dynamic changes of the UWB channel and improve the accuracy of signal estimation and detection;

(4) The system structure is simple, the operation complexity is low, and the detection time is short. In occasions where the detection accuracy is not high, the feature extraction results after nonlinear transformation can be directly used for judgment. At this time, the system structure is simpler and the detection time is shorter.

Description of drawings

Fig. 1 is a block diagram of the method of signal estimation and detection based on nonlinear transformation in the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

Given two assumptions, H ₀ - there is only additive Gaussian noise n(t) in the surrounding environment; H ₁ - there are allocated (licensed) wireless spectrum signals x(t) and additive Gaussian noise n(t) in the surrounding environment t), namely

H ₀ : y(t)=n(t)

H ₁ : y(t)=x(t)+n(t)

Let the received signal be y(t), 0≤t≤T.

Now estimate and detect whether there is a signal x(t) in the received signal y(t), the implementation steps are shown in the block diagram in Figure 1, and the details are as follows:

First, a nonlinear transformation is performed on the received signal y(t). One of the simplest nonlinear transformations is

f(t)=y ² (t)

The above-mentioned nonlinear transformation process is actually a square process, and a squarer is used.

It can also be transformed by other nonlinear transformers, such as a typical nonlinear transformer:

f(t)＝ay ² (t)-by(t-τ ₁ )y(t+τ ₂ )-cy(t-τ ₃ )

Where a, b and c are weighting coefficients, and τ ₁ , τ ₂ and τ ₃ are delay correlation times respectively.

Perform time-frequency transformation on the non-linearly transformed signal f(t)

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; j</span>}\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;ft</span>}}\mathrm{<span\; class="notranslate">\; dt</span>}$

The above-mentioned time-frequency transformation process is realized by using a fast Fourier transformer.

Take the modulus of F(f) after time-frequency transformation, and obtain the frequency f ₀ and 3dB bandwidth B corresponding to the maximum value accordingly, as the priori spectrum characteristics of the signal x(t), that is, priori knowledge.

According to the prior knowledge obtained above, the received signal is matched and filtered, and the filtered signal s(t) is output

$\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; d\&tau;</span>}$

The matched filter function h(t) is determined by the prior knowledge f ₀ , that is, h(t)=cos[2πf ₀ (Tt)]+sin[2πf ₀ (Tt)], where τ is the delay time.

According to the signal F(f) after time-frequency transformation and the coefficient α and bandwidth B set according to the false alarm rate and missing alarm rate required by the system, the decision threshold setter sets the decision threshold η

η=α|F(f)| ² /B

The decision unit performs hypothesis (H ₀ and H ₁ ) testing on the matched filter output s(t) according to the set decision threshold η.

If H ₀ is established, the received signal y(t)=n(t), which does not contain signal x(t), and there is no signal x(t) in the corresponding frequency band; if H ₁ is established, the received signal y(t) )=x can use this frequency band for communication (t)+n(t), then there is a signal x(t) in this frequency band, that is, the allocated (authorized) wireless spectrum signal x(t) is using this frequency band, so This frequency band cannot be used for communication.

Claims (4)

1. estimate and detection method that based on the signal of nonlinear transformation the wireless frequency spectrum signal that surrounding environment is existed carries out two kinds of hypothesis: do not have wireless frequency spectrum signal H ₀There is wireless frequency spectrum signal H ₁

H ₀：y(t)＝n(t)

H ₁：y(t)＝x(t)+n(t)

N (t) representes additive Gaussian noise, the wireless frequency spectrum signal that x (t) expression has been authorized, and the signal that receives is y (t), whether 0≤t≤T to existing signal x (t) to estimate among the signal y (t) and detecting, is characterized in that comprising the following steps:

1) nonlinear transformation is carried out nonlinear transformation with the broadband signal y that receives (t) through the non-linear converter, obtains nonlinear transformation signal f (t);

2) correlated characteristic of signal is extracted in feature extraction the signal f (t) after nonlinear transformation, and offers matched filter as priori;

3) matched filtering detects, and matched filter is according to step 2) in the priori that draws said reception signal y (t) is carried out matching detection, coupling output result is as the foundation of judgement;

4) judgement, decision device are judged the matched filter testing result according to the decision threshold of setting judge whether there is signal x (t);

Said step 2) obtaining of the said priori in is:

1) the signal f (t) after the nonlinear transformation is carried out time-frequency conversion through fast Fourier transformer and obtain time-frequency conversion signal F (f)

$F\left(f\right)={\&Integral;}_0^{T}f\left(t\right)e^{-j2\mathrm{\&pi;ft}}\mathrm{dt};$

2), and obtain the pairing frequency f of maximum in view of the above to the F behind the time-frequency conversion (f) delivery ₀With three dB bandwidth B, as the priori spectrum signature of signal x (t), i.e. priori.

2. the signal based on nonlinear transformation according to claim 1 is estimated and detection method, it is characterized in that the non-linear converter in the said step 1) does

f(t)＝ay ²(t)-by(t-τ ₁)y(t+τ ₂)-cy(t-τ ₃)

Wherein a, b and c are weight coefficient, τ ₁, τ ₂And τ ₃It is respectively time-delay correlation time.

3. the signal based on nonlinear transformation according to claim 1 is estimated and detection method; It is characterized in that said matching detection in the said step 3) be according to said priori through matched filter to received signal y (t) carry out matched filtering, output filtering signal s (t)

$s\left(t\right)={\&Integral;}_0^{T}y(t-\mathrm{\&tau;})h\left(\mathrm{\&tau;}\right)\mathrm{d\&tau;}$

Matched filter function h (t) is by priori f ₀Decision, i.e. h (t)=cos [2 π f ₀(T-t)]+sin [2 π f ₀(T-t)], wherein τ is a delay time.

4. the signal based on nonlinear transformation according to claim 1 is estimated and detection method, it is characterized in that said step 4) decision threshold is set at

η＝α|F(f)| ²/B

Wherein alpha will be set according to the false alarm rate and the false dismissed rate of system requirements,

Decision device detects said output filtering signal s (t) according to above-mentioned decision threshold η, if H ₀Set up, receive signal y (t)=n (t), wherein do not contain signal x (t), can use y (t) corresponding frequency band to communicate; If H ₁Set up, contain signal x (t) in y (t) corresponding frequency band, therefore cannot use the corresponding frequency range of y (t) to communicate.

Images