JPS6088372A - Signal detector - Google Patents

Abstract

PURPOSE:To enable high speed detection, by applying the treatment of a matched filter by using a future value due to an estimate. CONSTITUTION:Data Xk subjected to A/D conversion by an A/D converter 1 is sent to data buffer 2 storing data from the present time to the past only over a P-1 cycle. A MEM-AR model estimation device 3 forms a self-regression model according to a max. entropy method. A matched filter 4 inputs the output 7 of the data buffer and an estimate valve 8 to output Yk. A discriminator 5 performs detection from Yk. By applying the treatment of the matched filter as mentioned above, high speed detection is performed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] This invention pertains to signal detectors for finding significant signals buried in noise.

[Prior art]

FIG. 1 shows the principle of a conventional signal detector using a matched filter. Let x(t) be the signal of interest, and let x(t) be the digital signal at each sampling period Ts. As shown in FIG. 1(a), a block-shaped significant component appears in Xk. Let k be the instantaneous time, and consider a matched filter as shown in FIG. 1(b) which has a shape very similar to the significant component, and its output yk is given by: yk is x
After the peak of the significant component of k appears, the peak occurs with a delay of about half the duration of the waveform of the significant component, as shown in FIG. 1(c). This delay time was a major hindrance to platforms intended for instantaneous detection.

[Summary of the invention]

In order to eliminate this drawback, the signal detector according to the present invention has been devised as described below. An AR (autoregressive) model wpMEK (maximum entropy method) is fitted to the signal Xk. The details will be discussed later. If k is the current time according to the AR model...

xk-2, Xk-1, xk are expressed as (n=k, k-1, k-2, . . .). Therefore, the predicted value of the future value xk+1, xk
+2,... and multiply by these. In other words, since the future signal can be used at the current time as shown in Fig. 2(a), the matched filter can be used as shown in Fig. 2(b).
) can be obtained by applying it to future data. Here, L is an integer of 1≦L≦N-1 (6). Therefore, YK differs from yk in Figure 2 (
As shown in C), it is possible to make the peak occur almost at the same time as the peak of the significant component in xk. As described above, this signal detector is characterized by having high-speed detection capability based on AR model prediction.

[Embodiments of the invention]

FIG. 3 shows the implementation of the device d according to the invention.

In the figure, (1) is an A/D converter, (2) is a data buffer that stores A/D converted data for P-1 periods past the present, and (3) is an AR model created by MEM. MEM-AR model predictor for predicting future values, (4
) is a matched filter that inputs the data buffer value and the predicted value, and (5) is a discriminator that performs detection from the output Yk of matched filter (4). Also, (6) is the value xk-(p-1), ..., xk in the data buffer
, (7) is also xk+L-(N-1),...,x
k, and (8) is xk+1, xk+2, ..., x
k+L, and dk represents the detection result. Here, P is an integer satisfying M<P(7).

In the MEM-AR model predictor, the signal (6), i.e. uj
=xk-P+j(8) (j=1.2...,P) Find the coefficients a1, a2,..., aM(9) of the AR model as follows.

Intermediate variable bmj (m=0, 1,..., M:j=1, 2,...
, P-m) (10) b'mj (m=0, 1,...M:j=1,2,...
, P-m) (11) amj (m=0, 1,...M; j=0, 1,...,
m) (12) Pm (m=0, 1, . . . , M) (13) is defined as follows.

First, b0j=b'0j=uj(j=1.2...,P)(
14) b1j=b'1j=uj+1 (j=1, 2...
P-1) (15) am0=-1 (all m) (17) amj=0 (all m<j) (18). m here.

Let's increase m = 1, 2, . . . , m (19), and first increase the amm. Next, amj=am-1,j-am,m,am-1,m-j(
20) Find amj (j=0, 1,..., m) by (j=0, 1,..., m). Next, bmj=bm-1j-am-1,m-1b'-1,j(
22) b'mj=b'm-1,j+1-am-1,m-
1bm-1, j+1 (23) (j=1, 2,, p-m), bmj, b'mj (j=1, 2,..., p-m
). Returning to equation (20) again, increase the value of m by one and repeat the same steps, bmj, b'mj. Add amj. Also, pm is pm=pm-1(1-am2m)(24)(m=1,2
,...,M) (25).

aM,j (j=1, 2,..., obtained in this way)
Autoregressive model coefficients are obtained from M) by aj=aMj (j=1, 2, . . . M) (25).

Using the coefficients a1, a2, . . . , aM determined in this manner, xk+1, xk+2', .

Matched filter (4) calculates output Yk using equation (5). The discriminator (5) calculates the detection result based on the output of the matched filter (4). Here, δ is an appropriate value.

〔Effect of the invention〕

In this way, high-speed detection can be performed by performing matched filter processing using predicted future values.

Although this invention has been described as a device for detecting a significant component in a signal from a magnetic sensor, it can also be used for target detection processing in radar or the like.

[Brief explanation of the drawing]

Figure 1 is a diagram showing the effect of a conventionally used matched filter for signal detection, Figure 2 is a diagram showing the effect of a matched filter using prediction by MFM-AR model prediction, and Figure 3 is a diagram showing the effect of the present invention. FIG. 3 is a diagram showing an example of a signal detector. In the figure, (1) is an A/D converter, (2) is a data buffer, (3) is a MEM-AR model predictor, (4) is a matched filter, and (5) is a discriminator. In the drawings, the same or corresponding parts are designated by the same reference numerals. Agent Masuo Oiwa Figure 1 Figure 3

Claims (1)

[Claims] An A/D converter converts this signal into a digital signal by taking all input signals including block-shaped significant components buried in noise obtained from magnetic sensors, etc., and converting this digital signal into a digital signal with a period equal to the conversion period of the A/D converter. The maximum entropy method (
Auto Regreseive Model (Maxmum Eutropy Method)
a maximum entropy autoregressive model predictor that adapts the maximum entropy autoregressive model predictor; a matched filter that combines the output of the maximum entropy autoregressive model predictor with the value stored in the data buffer (as input);
1. A signal detector comprising a discriminator for detecting the presence of a significant component from the output of a matched filter, and capable of detecting the presence of a block-shaped unexpected component at an intermediate time.