JPH0310469A - Ghost elimination device - Google Patents

Abstract

PURPOSE:To reduce the arithmetic required time for nearby ghost detection by applying Fourier transformation and inverse Fourier transformation with a low degree to a time serial signal segmented from a nearby area and less sampling point number. CONSTITUTION:A nearby ghost elimination section 31 and a normal ghost elimination section 33 are connected in cascade and an original television signal being an object of ghost elimination fed to an input terminal IN is fed to a nearby ghost detection section 32 and a normal ghost detection section 34. The nearby ghost detection section 23 applies Fourier transformation to a digital signal extracted from a nearby ghost area of a received television signal, divides a Fourier coefficient group of a ghost detection reference waveform stored in advance with the Fourier coefficient group, applies inverse Fourier transformation to generate a time serial signal group, extracts the nearby ghost area of the time series signal group and supplies the result to a pseudo nearby ghost generation transversal filter 31a as the tap coefficient group. Thus, no much time is required for the calculation of the nearby ghost processing system having lots of step numbers.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a ghost removal device installed in a television receiver.

(Prior Art) In general, a ghost removal device includes a transversal filter for generating pseudo ghosts, a ghost detection unit that detects the occurrence of ghosts and sets a group of tap coefficients of the transversal filter, and a transversal filter. The signal combining section removes the ghost contained in the original television signal by combining the pseudo ghost generated by the filter and the original television signal.

These ghosts are roughly classified into close ghosts and non-close ghosts depending on the degree of proximity to the original signal. In other words, as shown in Figure 3, the original signal is converted into a waveform (A
), a nearby ghost causes waveform distortion illustrated by the solid line portion α, and a non-proximal ghost causes waveform distortion illustrated by the solid line portion β. Proximity ghosts include various factors related to transmission characteristics that cause waveform distortion, whereas non-proximity ghosts are based on a phenomenon unique to ghosts, which is the formation of detour propagation paths. For this reason, non-proximity ghosts are commonly referred to as ghosts.

As shown in Figure 3, ghosts usually appear at a distance from the original signal on the time axis, and as the ghosts are removed, new child ghosts and grandchild ghosts are created, gradually attenuating at predetermined time intervals. They tend to appear repeatedly.

Therefore, when removing this non-proximate ghost, it is desirable to perform cyclic synthesis of the pseudo ghost generated from the original television signal and the original television signal.

On the other hand, since proximity ghosts appear overlapping the original television signal on the time axis, it is necessary to delay the original signal.
Cyclic configuration cannot be applied.

Therefore, the removal of proximal ghosts and non-proximal ghosts is not done by a single ghost removal device, but by connecting dedicated ghost removal devices in tandem, which first removes proximal ghosts and then removes non-proximal ghosts. It is desirable to divide the process into two stages.

1988 patent IJ16756 related to the applicant's earlier application
No. 0 discloses a ghost removal device having the configuration shown in FIG.

This ghost removal device includes a proximity ghost removal section 31, a proximity ghost detection section 32, a normal ghost removal section 33, and a normal ghost detection section 34. It is configured to perform removal processing and output from the output terminal OUT.

As shown in FIG. 5, the proximity ghost removal section 31 includes:
Transversal filter 3 for pseudo-proximity ghost generation
1a, a delay circuit 31b, and an adder circuit 31C.

Further, as shown in FIG. 6, the normal ghost removal section 33 has a cyclic configuration consisting of a pseudo normal ghost generation transversal filter 33a and an adder circuit 31b. The pseudo-ghost generation transversal filters 31a and 33a, as shown in FIG.
b42G...42n, and multipliers 43a, 43b, 43 that multiply the signals drawn from each delay device by tap coefficients.
c...43n and an adder 4 that combines the outputs of each multiplier.
It is composed of 4.

The configuration in which the entire ghost detection area is divided into a nearby ghost detection area and a normal ghost detection area, and each is detected and removed in stages is based on the following principle.

That is, the reference waveform for ghost detection (OCR waveform) extracted from the received television signal and the time-series signal sequence before and after it are g (t), and its frequency spectrum is G (j
ω), and the transfer characteristic of an ideal ghost removal device is H(jω), and the frequency spectrum of the reference waveform for ghost detection before being distorted by ghosts is R.
(jω), then G(jω) x H(jω) = R(jω) (1).

From equation (1), the transfer characteristic H of the ideal ghost removal device
(jω) becomes H(jω) = R(jω)/G(jω) (2).

G(jω) can be obtained by Fourier transforming the impulse response. For example, as a reference waveform,
It can be easily determined by using a pulse waveform of inx/x. In addition, since the 5inx/x signal and the vertical synchronization signal can be treated as step responses, they can be converted to impulse responses by differentiating them, and G(j
ω) can be obtained.

Transfer characteristic G(jω) up to just before the ghost removal device
(near the vicinity of the reference waveform)? The fJf area and the normal ghost area that appears after this proximity ghost [1 area] are divided, the former being Gl (jω) and the latter being G2 (jω).
shall be. That is, G (jω) = 01 (jω) + 02 (jω)... (3
).

Using equation (3), equation (2) becomes: H(jω) = R(jω)/(Gl(jω) 10G2 (jω)
] = (R (jω) / Gl (jω)) / (1 + G2 (jω) / Gl (jω)] ・ ・ ・
(4) It is expressed as

Therefore, the transfer characteristic H(jω
) is Hl (jω) = (R(jω)/Gl(jω)] ・ ・ ・ (5) H2(jω) = 1/ [(1+G2 (jω)/C1(jω)))
・ ・ ・ (6) The transfer function H1(jω) of the proximity ghost removal unit is
It is usually divided into two parts: the transfer function H2(jω) of the ghost removal section.

(5) According to 2, the transmission amount BHI of the proximity ghost removal unit
(jω) is the transfer function Gl (jω
) and the ideal baseband characteristic R(jω). Further, as the tap coefficients of this transversal filter for removing nearby ghosts, a value obtained by inverse Fourier transform of Hl (jω) may be used.

Furthermore, according to equation (6), the transfer function H2 (jω) of the normal ghost removal unit is obtained from the transfer characteristic Gl (jω) of the proximity ghost removal device and the transfer function G2 (jω) of the normal ghost) 614 region. It will be done. Further, as the coefficients of this normal ghost removal transversal filter, a value obtained by inverse Fourier transform of -G2(jω)/Gl(jω) may be used, and it may be constructed in a cyclic type.

In this way, the coefficients of the ghost removal filter from close range to normal can be found just by knowing the signal input to the ghost removal device.

The tap coefficient group to be set in the transversal filter for pseudo-proximity ghost generation calculated from equation (5) generally becomes an infinite-length impulse response. Since the number of tap stages of an actual filter is finite, the waveform finally obtained differs from the theoretical value, and residual ghosts that cannot be completely removed occur. Regarding this residual ghost, there is no problem in practice if the pseudo-proximity ghost generation transversal filter has an appropriately large number of tap stages for the proximal ghost region to be removed. When the number of tap stages of the transversal filter for generating pseudo-proximity ghosts is insufficient, residual ghosts become a problem, but this problem can be solved as follows.

In other words, the transfer characteristic H1 of the ideal proximity ghost removal section
(jω) is Hl (jω) = R(jω)/Gl(jω)・(
7) However, the transfer characteristic C1 (jω) of the proximal ghost remover that is actually used is CI (jω) = 81 (jω) −E Nω) (8
), the error E(jω) with respect to the ideal value H1(jω)
) exists. Therefore, with respect to the overall transfer characteristic G(jω) including ghosts, the transfer characteristic after removing the nearby ghosts is given by Equations (3), (7), and (8), as follows: G(jω)・C1 (jω) = (Gl (jω)+G2 (jω)]X()+1
(jω) −E (jω)]=R(jω)+G2 (j
ω)・Hl (jω) (Gl(jω) 10G2 (
jω))E(jω)... (9) It becomes.

If the transfer characteristic of the normal ghost removal section is C2(jω), the final transfer characteristic after normal ghost removal is R(jω).
Therefore, G(jω) ・C1(jω) ・C2(jω)=R(j
ω) ・ ・ ・ ・ The relationship (10) is obtained.

From equations (9) and (lO), the transfer characteristic C2(jω) of the normal ghost removal section is: C2(jω) = R(jω)/(G(jω)・C1(jω)) = R(j
ω)/(R(jω)+G2 (jω)Hl (jω)
-(Gl(jω)+02 (jω)]XE (jω)] -(1+G2 (jω)/Gl(jω)E(jω)(
Gl (jω) + G2 Nω)]/R(jω)] ・ ・ ・ (11)

From equation (11), the tap coefficient group of the recursive pseudo-normal ghost generation transversal filter is −G2 (jω)/Gl(jω) +E(jω) (Gl (jω)+G2 (jω
)〕/R(jω)〕 ・ ・ ・
An inverse Fourier transform of (12) may be used.

Therefore, the processing contents of the nearby ghost detection section 32 and the normal ghost detection section 34 in FIG. 4 are as shown in FIG. 8.

First, in the proximity ghost detection section 32, the OCR signal g obtained by converting the reference waveform for ghost formation into a digital signal is detected.
(t) is extracted (step 51).

Using the window shown in FIG. 9(A) from this OCR signal g([), a time-series signal gl of a nearby ghost area is obtained.
(t) is cut out (step 52). However, the 9th
The upward arrow drawn in the figure indicates the insertion position of the OCR signal. Next, a Fourier coefficient group Gl (jω) is calculated by Fourier transformation of gl (t) (step 53), and H is calculated by dividing R(jω)/Gl(jω).
l(jω) is calculated (step 54), and this calculated value is subjected to inverse Fourier transform to generate a time series signal hl(t) (step 55), which has the characteristics shown in FIG. 9(C). Proximity filter coefficients CI(t) are extracted using a window (step 56) and provided to a transversal filter for generating pseudo-proximity ghosts.

On the other hand, an error signal e(1) due to the finite number of tap stages is extracted from the time-series signal hl(t) using a normal region window with characteristics as shown in FIG. 9(D>). 56) is Fourier-transformed and is normally supplied to the ghost detection unit 34 as a Fourier coefficient group E(jω) of the error signal.This supply path is formed between the output terminal EO and the input terminal Ei in FIG. There is.

On the other hand, the normal ghost detection section 34 performs OCR signal extraction processing (step 51) which is shared with the proximity ghost detection section 32.
), the time-series signal g (t) is extracted in Fig. 9 (
A time-series signal g2 (t) in the normal region is cut out using a window with the characteristic shown in B) (step 61), and is subjected to Fourier transformation to generate a Fourier transform group G2 (jω) (step 62). In the next step 63, the Fourier coefficient group of equation (12) is generated, subjected to high-level suppression by nonlinear processing (step 64), inverse Fourier transformed (step 65), and usually transversal for ghost removal. fed to the filter.

(Problems to be Solved by the Invention) In general, the nearby ghost detection area is sufficiently narrower than the normal ghost detection area, so the number of sampling points cut out using each window is sufficiently smaller in the former. Therefore, the order of Fourier transform in step 53 in FIG. 8 should be sufficiently smaller than the order of Fourier transform in step knob 62. However, in the processing procedure of FIG.
When calculating 2(jω)/Gl(jω), it is necessary to match the orders of the divisor and dividend. As a result, step 53
During the Fourier transform in step 62, zero amplitude sampling points over the normal region are added to the time series signal gl(t) cut out from the adjacent region, and processing is performed to match the order to the order of the Fourier transform in step 62.

As a result, there is a problem in that it takes time for the processing system to calculate the proximity ghost, which requires a large number of steps, and the ability to follow the ghost, which changes relatively quickly as a moving object approaches or moves away, deteriorates.

(Means for Solving the Problems) According to the ghost removal device of the present invention, the proximity ghost detection unit that sets a tap coefficient group to the pseudo proximity ghost generation transversal filter of the acyclic proximity ghost removal unit is configured to: Proximity ghost in received television signal? Fourier transform the digital signal extracted from the J region, divide the Fourier coefficient group of the reference waveform for ghost detection held in advance by this Fourier coefficient group, and then perform inverse Fourier transform to generate a time series signal group. The apparatus includes means for extracting a proximal ghost region of a time-series signal group and supplying the extracted proximal ghost region to the pseudo proximal ghost generation transversal filter as a group of tap coefficients.

In addition, the normal ghost detection unit sets a group of tap coefficients in the pseudo-normal ghost generation transversal filter that constitutes the recursive normal ghost removal unit. The signal is Fourier transformed, the product of this Fourier coefficient group and the Fourier coefficient group of the time series signal group extracted by the nearby ghost detection unit is calculated, and the Fourier transform of this product and nI of the reference waveform for ghost detection held in advance is calculated. The apparatus is provided with means for calculating a difference with a group of coefficients, subjecting this difference to inverse Fourier transform to generate a group of time-series signals, and supplying this to a pseudo-normal ghost generation transversal filter as a group of tap coefficients.

(Function) If we extract the signal in the proximal ghost region from the time series signal h (t) obtained by performing inverse Fourier transform on the above-mentioned equation (2) (t).

In addition, by cutting out the time-series signal gl (L) in the proximity region from the 8 time-series signals of the OCR signal, the tap coefficients of the transversal filter for pseudo-proximity ghost generation are calculated based on low-order Fourier transform and inverse Fourier transform. Can calculate groups.

The frequency spectrum Y(jω) of the OCR signal from which the proximity ghost has been removed and which is normally supplied to the ghost removal unit is determined from the transfer characteristic HL (jω) obtained by Fourier transforming the impulse response h1(t) of the proximity ghost removal unit. Y(jω) −〇 (jω)・Hl (jω)・・
It is calculated as (l 3).

Therefore, the transfer characteristic 1(2(jω)) of the normal ghost removal unit for equalizing Y(jω) to R(jω) is: H2(jω) = R(jω)/Y(jω) -1/( 1-X (jω)] ... (14)
However, X(jω) = l−G(jω)・Hl (jω)/R(jω)・
・(15) becomes.

Therefore, the time-series signal group obtained by inverse Fourier transform of X(jω) in equation (15) is used as the tap coefficient group of the transversal filter for generating pseudo-proximity ghosts in the normal ghost removal section configured as a cyclic type. Just set it.

In reality, considering that the band of R(jω) is finite, the band of the above X(jω) is limited by R(jω), and X(jω) #R(jω) −G (jω)・Hl ( jω)・・
- The time series signal group obtained by inverse Fourier transform of (16) may be set as the tap coefficient group of the pseudo-normal ghost generation transversal filter.

In this way, G(jω) is generated by processing the proximity area and the normal area in the pseudo-normal ghost detection unit, and by changing the algorithm to take Hl (jω) from the proximity ghost detection unit, It becomes possible to significantly reduce the order of Fourier transform and inverse Fourier transform of the ghost detection unit, and to significantly shorten the time required for calculation.

Hereinafter, the present invention will be explained in more detail with reference to Examples.

(Embodiment) A ghost removal device according to an embodiment of the present invention is similar to the conventional device shown in FIG. [The original television to be ghost removed supplied to N is the proximity ghost detector 32 and the normal ghost detector 3
It is configured to be supplied to 4. Further, in the proximity ghost removal section 31, similar to the conventional device shown in FIG. 5, a pseudo proximity ghost generation transversal filter 31a, a delay section 31b, and an addition section 31C are connected in an acyclic manner. It is composed of: Further, the normal ghost removal section 33 is also configured by connecting a pseudo normal ghost generation transversal filter 33a and an addition section 33b in a cyclic manner, similar to the conventional device shown in FIG. . Both of the transversal filters 31a and 33a, similar to the conventional device shown in FIG. It consists of an adder that combines the outputs of the group of devices.

The processing contents of the near ghost detection section and the normal ghost detection section will be explained with reference to the flowchart shown in FIG. 1 and the window characteristic diagram shown in FIG. 2.

First, in the proximity ghost detection section, a time-series signal g([) over the proximity region and normal region of the reference waveform for ghost detection (OCR waveform) included in the received television signal that has been converted into a digital signal is extracted. (Step 1
1). Next, the time-series signal g1 in the nearby region is cut out using a window with characteristics as shown in FIG. 2(A) (step 12), and a group of Fourier coefficients CI (jω) is generated by Fourier transformation. (Step 13).

Since the number of sampling points of the time-series signal is small, a small value is sufficient for the order of this Fourier transform, and typically 12
At the 6th order, which is about order a, division with the Fourier coefficient group R(jω) of the undistorted reference waveform held in advance is performed to generate the coefficient group H1(jω) (step 14). An inverse Fourier transform of order 128 is performed on the coefficient group H1(''ω) to generate a time series signal hl(L) (step 15). This time series signal h1 (t) is the second
Using the window with the characteristics shown in figure (C), the time series signal c is
l(t) (step 16), and is supplied as a group of tap coefficients to a transversal filter for generating pseudo-proximity ghosts in the proximal ghost removal section. Also,
In step 16, the time series signal k(1) cut out by the window with the characteristic shown in FIG. is added, and the typical equation is subjected to Fourier transform of order 1024 (step 17). This Fourier coefficient group K(jω)
is normally supplied to the ghost detector. This supply route is
It is formed between the output terminal EO and the input terminal Ei in FIG.

On the other hand, in the normal ghost detection section, the time-series signal g (t
is extracted, and a nearby region and a normal region are cut out using a window with the characteristics shown in FIG. 2(B) (step 2).
1). Since this time-series signal g(t) includes a large number of sampling points, it is typically Fourier-transformed at an order of 1024 (step 22), and is transformed into a Fourier coefficient group G(jω).
It is passed to processing step 23 as . Processing step 23
Now, the Fourier coefficient group G(jω
) and the Fourier coefficient group H1(jω) that has passed through the processing step 17 of the nearby ghost detection unit, and this product is calculated as the Fourier coefficient group R(j) of the undistorted OCR held in advance.
By subtracting from ω), we can obtain the above-mentioned (16) 2 of X(
jω) is generated.

This Fourier coefficient group supplied to

As described above, from the time series signal k (t), the Fourier coefficient group K (j
An example of a configuration in which Fourier transform for generating ω) is performed within the proximity ghost detection unit is illustrated. However, if this Fourier transformation is normally performed within the ghost detection section, the calculation time can be further reduced.

(Effects of the Invention) Since the ghost removal device of the present invention has the above-described configuration, it is possible to perform low-order Fourier transform and inverse Fourier transform on a time series signal with a small number of sampling points extracted from a nearby region. This greatly reduces the time required for calculations to detect nearby ghosts.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing the procedure of processing performed in a nearby ghost detection unit and a normal ghost detection unit that constitute a ghost removal device according to an embodiment of the present invention, and FIG. A conceptual diagram showing the characteristics of various windows used for extraction, FIG. 3 is a waveform diagram for explaining the concept of close ghosts and normal ghosts, and FIG. 4 shows the configurations of the conventional device and the ghost removal device of the above embodiment. 5 and 6 are block diagrams showing the configurations of the proximity ghost removal section 31 and normal ghost removal section 33 of the ghost removal device shown in FIG. 3, and FIG.
FIG. 8 is a block diagram illustrating the configuration of the transversal filter for pseudo ghost generation shown in FIG. The figure is a conceptual diagram showing the characteristics of various windows used for cutting out time-series signals in the process of FIG. 8. IN... Input terminal for television signal to be ghost removed, 31... Proximity ghost removal unit, 31a... Transversal filter for pseudo proximity ghost generation, 3
2... Proximity ghost detection section, 33... Normal ghost removal section, 33a... Transversal filter for pseudo normal ghost generation, 34... Normal ghost detection section.

Claims (1)

[Scope of Claims] A nearby ghost detection unit that detects a nearby ghost appearing in the vicinity of a reference waveform for ghost detection included in a received television signal and generates a group of tap coefficients to be set in a transversal filter for generating a pseudo nearby ghost. and a normal ghost detection unit that detects a normal ghost that appears after the near ghost and generates a group of tap coefficients to be set in a transversal filter for generating a pseudo normal ghost; and the transversal filter for generating a pseudo normal ghost; an acyclic proximity ghost removal unit comprising a synthesis means for combining the generated pseudo-proximity ghost with a television signal to be ghost-removed; a transversal filter for generating the pseudo-normal ghost; and a transversal filter for generating the pseudo-normal ghost; A ghost removal device comprising: a cyclic normal ghost removal section including a synthesis means for combining a ghost with a television signal to be ghost removed output from the proximity ghost removal section, wherein the proximity ghost detection section includes A time series signal is obtained by performing Fourier transform on the digital signal extracted from the nearby ghost region of the received television signal, dividing the Fourier coefficient group of the reference waveform for ghost detection held in advance by this Fourier coefficient group, and performing inverse Fourier transform. the normal ghost detection unit includes means for generating a group of tap coefficients, extracting a nearby ghost region of the time-series signal group, and supplying the same to the pseudo nearby ghost generation transversal filter as a group of tap coefficients; The digital signals of the nearby ghost region and the normal ghost region extracted from the signal are subjected to Fourier transform, and the product of this Fourier coefficient group and the Fourier coefficient group of the time series signal group extracted by the nearby ghost detection section is calculated, and this product is and the Fourier coefficient group of the reference waveform for ghost detection held in advance is calculated, and this difference is inverse Fourier transformed to generate a time series signal group, which is applied to the pseudo-normal ghost generation transversal filter. A ghost removal device characterized by comprising means for supplying tap coefficients as a group.