JPH0511448B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a processor for removing impulsive noise caused by ignition noise, motor noise, etc. contained in an audio signal of an in-vehicle radio receiver.

[Prior Art] A method for removing an impulsive noise signal includes a method of detecting the start of the pulse noise, holding the signal at a value before the start of the pulse noise for a certain period of time, and smoothing the signal. Referring to FIG. 9, a indicates a signal containing noise, S indicates a signal, and N ₁ to _{N 3} indicate noise.
These noises are detected, a gate pulse like c with a constant width t ₀ is generated from the time of detection, and the amplitude of the signal is kept at the previous value during the pulse width period of this gate pulse to eliminate the noise like b. Obtain the removed signal.

[Problem that the invention seeks to solve]

However, in this method, since the gate pulse width t ₀ is constant, some noise with a wide width such as N ₂ cannot be completely removed and remains. To avoid this, if the gate pulse width t ₀ is increased, The flat portion that occurs in the area where noise is removed becomes large, and the distortion of the signal waveform becomes large, resulting in the possibility that noise removal becomes noise addition. In addition, the timing of the generation of the gate pulse must correctly match the noise N ₁ , N ₂ , ..., and if this happens, incomplete noise removal will occur frequently, and non-noise parts of the signal will be removed, increasing distortion. .

The present invention improves these points by using digital signal processing technology to eliminate or reduce impulsive noise caused by igniter operation and motor commutator brushes specific to automotive receivers, and achieve the desired noise suppression. It is an attempt to obtain characteristics.

[Means for solving problems]

The present invention provides a processor for removing impulsive noise contained in an audio signal, which includes a detector that detects noise contained in the audio signal and outputs a detection signal whose pulse width is the duration of the noise; The signal is sampled at a frequency of ₁ , each obtained sample is digitized, the samples for noise sampled during the generation period of the detection signal output from the detector are removed, and the samples for the noise are compared with the samples before and after. a first circuit that interpolates and replaces based on the samples; and resamples each sample output by the circuit at a frequency p;
The resampled sample is digitally low-pass filtered, and the result is resampled at a frequency _{of 2} (where p=M ₁ =N ₂ , M and N are integers and M
>N) A signal processor and a second circuit that converts the output of the processor into analog and further performs low-pass filtering.

[Effect]

In the present invention, impulse noise is removed by interpolating noise intervals in the process of converting the sampling frequency system, such as by resampling the audio signal or thinning out the output of a digital low-pass filter. Referring to FIG. 4, a indicates each sample obtained by sampling the audio signal at frequency ₁ , and 0, 1, 2, . . . are the 0th, 1st, 2nd, . . . samples. In this example, the third sample is abnormal, and is a sample of noise N. The present invention resamples such samples at frequency p. p is synchronized with ₁ and p=E ₁ . Here, E is an integer, and in this example, E=3. When resampling at such a frequency, as shown in b, samples 0, 1, 2, ... (in b, 0, 3, 6, ...) at ₁ and each of the two samples 1, 2, ... 4, 5,...
(These are all 0) are obtained. Separately, the audio signal is passed through a high-pass filter to detect noise N, and the sample at the time when noise N is detected is p
It is not included in the sample at , and the sample for this part is generated separately as the average value of the samples before and after it. Sample 9 of b shows this average value. Note that if the signal level is low, it may be possible to simply set it to 0 instead of using the average value.

Next, the sample sequence 0, 1, 2, 3,... shown in b
... is passed through a digital low-pass filter to obtain a sample sequence c. In other words, when passing through this filter, samples 1, 2, 4, 5, etc., which were 0 before passing, are passed through this filter.
As shown in the figure, is reproduced to the value at the relevant timing in the low frequency determined by samples 0, 3, 6, . Further, in a sample such as sample 9, which is not a sample of an audio signal, the shape is such that the low frequency signal determined from the sample rows on the left and right sides of the sample are matched, and the envelope waveform is discontinuous. Such a sequence of samples is then resampled at frequency ₂ . Frequency ₂ is synchronized with p, and there is a relationship p=F ₂ . Here, F is an integer, and in this example, F=2. In this way, the digital low-pass filter outputs 0, 1,
In this example, 2, 3, ... can be extracted every other time and sampling at ₂ can be performed, and since it is partially extracted and the remaining part is discarded (in this example, half is discarded), this remaining part can be used to deal with noise. If a sample is included, it is possible to avoid including 9 in this example. By thus obtaining a sample sequence of d and applying it to an analog low-pass filter, it is possible to obtain an audio signal from which noise N has been removed.

FIG. 5 shows such processing in a block diagram. Sample the audio signal at frequency ₁ , resample it at frequency p, pass it through a digital low-pass filter,
The output is resampled at frequency ₂ and noise is removed in this process. Here, p is p=E ₁ = E ₂ (E>F)...(1) In other words, it is the least common multiple of ₁ and ₂ , and in this way, the sample points of ₂ are interpolated by the digital low-pass filter. It is possible to match the points, and each sample in the later stage can be captured more reliably than each sample in the earlier stage. ₁ is 40K for example
Hz, ₂ is 60KHz, p is 120KHz. If you have a sample value of 60KHz, _1/2 = 20KHz maximum with LPF of 20KHz
It is possible to restore the signal up to.

Next, to explain the filter, the frequency characteristic H(e ^j 〓) of a digital filter is generally expressed by the following formula.

H(e ^j 〓)= _∞ 〓 ^-∞ h(n)e ^-j 〓 ⁿ ……(2) h(n)=1/2π∫〓 _- 〓H(e ^j 〓)je ^jwn dω……(3 ) The unit impulse response is h(n)=1/2π∫〓 ^C _- 〓 _C (1)xe ^j 〓 ⁿ dω = Sin(ωCn)/πn=ωC/π×Sin(ωCn)/ω
Cn...(4). These are shown in FIG. As shown in the figure, equation (4) expresses the IIR (Infinite Impulse Response) function h(n).
Approximate to FIR (Finite Impuise Response) function hf(n).

hf(n)=h(n)Xω(n) ……(5) Let us consider configuring an FIR digital filter using equation (5) using N-1/2 discrete values on the left and right, centering on n=0 in FIG. 7b, as multiplication coefficients.

The cutoff frequency of the filter for restoring the discrete signal of the frequency ₁ system to an analog signal is
_1/2 , so the cutoff frequency ωc in Figure 6a is ωC=π...(7) Also, when a system with frequency ₁ is sampled at frequency p, ωC at the cutoff frequency _1/2 is ωC =π/E...(8). hd(n) in the above case is hd(n)=1/E×sinnπ/E/nπ/E (9), which is shown in FIG. Let hd(n), where the value of n in equation (9) falls within the range of equation (6), be the multiplication coefficient.
The FIR digital low-pass filter is shown in FIG. Now E=3, F=2 or p=3 ₁ = 2 ₂ ,
If ωC=π/3 and N=13, Figure 3 is obtained. The digital low-pass filter shown in FIG. 3c may be used.

〔Example〕

Fig. 2 shows the configuration of a car radio receiver to which the present invention is applied, and Fig. 1 shows its main part, that is, an embodiment of the present invention.The AM or FM radio receiver is as shown in Fig. 2. , a high frequency amplifier 10, a mixer 12, a local oscillator 14, an intermediate frequency amplifier 16, a discriminator (or detector) 18, and an audio amplifier 20, which amplify and select the broadcast waves received by the antenna 24.
The signal is detected and amplified, and the speaker 22 is made to sound. The inventive circuit 30 is inserted between the discriminator 18 and the audio amplifier 20 to eliminate impulsive noise. The detector 32 detects this noise. This detector 32 is equipped with a high-pass filter and a shaping circuit that converts the output into a rectangular wave, and generates a rectangular wave that is generated simultaneously with the pulse noise and whose pulse width is the duration of the noise.

FIG. 1 shows details of the circuit 30 of the present invention, including an analog low-pass filter 34 for band limiting, a sample hold circuit 36, an A/D converter 38, a signal processor 40, a D/A converter 50, and an analog circuit for interpolation. It consists of a low pass filter 52. Discriminator 1
The audio signal outputted by 8 passes through a low-pass filter 34 and enters a sample-and-hold circuit 36, where it is sampled at a frequency of ₁ . Each sample enters an A/D converter 38, where it is converted to a digital value, and the noise detection output from the pulse noise detector 32 removes (zeros) the sample for noise.

Each digitized sample is processed by signal processor 40 as described in FIG.
The processor 40 includes a RAM (random access memory) 42, an arithmetic circuit 44, a control circuit 46, and a ROM.
(read-only memory) 48, the coefficient hd(n) necessary to configure the digital low-pass filter is
Store it in ROM48. Each sample from the A/D converter 38 is written to the RAM 42 one after another,
The arithmetic circuit 44 calculates the sum of products with the coefficients of the ROM 48 . To explain with reference to FIG. 4, the output of the A/D converter 38 is for each sample 0, 1,
2. It is...(However, it is digitalized),
These are sequentially written into the RAM 42. The RAM write frequency is p, and each two in between are written as data 0, and the noise-responsive sample set to 0 is newly generated as the average value of the samples on its left and right, thus the data in Figure 4b is
Stored in RAM. As is clear from Figure 3, the product-sum operation is performed by multiplying 13 consecutive samples (N = 13 in this example) by the coefficients h _(-6) , h _(-5) ... h _(6). ,
This is done by finding the sum of these products, multiplying it by three, and repeating this process while shifting the entire sample at a pitch of one sample.

That is, Z ^-1 in Fig. 3 is a circuit that provides a delay corresponding to the time between each input sample x(n),
There are 12 samples, so 13 consecutive samples are the coefficients.
Simultaneously input to the multiplier M of h _(-6) , h _(-5) , ...h ₍₆₎ ,
After one sample time has elapsed, the rightmost sample disappears, a new sample enters the leftmost, the whole is shifted to the right by one sample, and a new group of 13 samples is simultaneously input to the multiplier M, and this process is repeated thereafter.
In order for the arithmetic circuit 44 to perform such processing, it reads 13 consecutive pieces of data from the RAM 42 and multiplies these with the coefficients h _(-6) , h _(-5) , ... h ₍₆₎ of the ROM 48. and the new 13 data shifted by 1 sample.
The data is read from the RAM 42 and the same process is repeated, and this is repeated thereafter.

As a result of such processing, each output 0, 1,
2, . . . are obtained, and these are written to the RAM 42. Then, if we take out every other product sum written to this RAM 42 (resampling it by ₂ ), we get the result shown in Figure 4d.
is obtained, which becomes the output of the signal processor 40.

The output of the signal processor is a D/A converter 50.
Convert to analog to obtain Figure 4e. This is (d)
Although they look the same, d is a digital signal and e is a discrete signal. 4e, the output of the D/A converter 50 is filtered through the low-pass filter 52.
, an analog audio signal with noise removed is obtained.

The delay circuit 54 delays the timing of signal processing to ensure that the samples at the output of the detector 32 are processed against noise. The control circuit 46 performs various controls.

〔Effect of the invention〕

As described above, according to the present invention, since the pulse noise section is removed using sampling frequency conversion, it is possible to provide a more natural and natural sound. Furthermore, by sampling the audio signal, pulse noise between discrete values is also removed at this processing stage, and an excellent noise removal processing effect can be expected.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a block diagram showing the configuration of a radio receiver to which the invention is applied, Fig. 3 is an explanatory diagram of a digital filter, and Fig. 4 is for explanation of operation. 5 is an explanatory diagram of the processing procedure of the present invention, FIGS. 6 to 8 are explanatory diagrams of a digital filter, and FIG. 9 is an explanatory diagram of conventional noise removal. In the drawing, 32 is a pulse noise detector, 36, 38
is a first circuit, 40 is a signal processor, 5
0 and 52 are the second circuits.

Claims (1)

[Claims] 1. A processor for removing impulsive noise contained in an audio signal, comprising: a detector that detects noise contained in the audio signal and outputs a detection signal having a pulse width equal to the duration of the noise; , sample the audio signal at a frequency of ₁ , digitize each obtained sample, remove samples for noise that were sampled during the generation period of the detection signal output from the detector, and convert the samples for noise into A first circuit that interpolates and replaces based on previous and subsequent samples, resamples each sample output by the circuit at a frequency p, performs digital low-pass filtering on the resampled sample, and filters the result at a frequency of _2.
(where fp=M ₁ =N ₂ , M, N
is an integer and M>N); and a second circuit that converts the output of the processor into analog and further performs low-pass filtering.