JP5103606B2 - Signal processing device - Google Patents

Description

The present invention is, for example, used to extend the number of bits of the quantized audio signal relates to a suitable signal processing equipment. Specifically, a good number of bits are expanded while blocking harmonic components due to the expanded bits.

In conventional signal processing equipment, to perform processing such as noise reduction using a feedback type of filter is known (e.g., see Patent Documents 1 and 2.).

  An apparatus for converting a bandwidth of a quantized signal using a feedback filter has also been proposed (see, for example, Patent Document 3).

That is, a device for processing a quantized signal using a feedback filter has been known.
US Pat. No. 6,053,318 US Pat. No. 6,154,547 US Pat. No. 3,889,108

  For example, in a compact disc (CD), it is generally widespread that an audio signal is digitized and recorded with a sampling frequency of 44.1 kHz and a quantization bit number of 16 bits. However, when the recorded 16-bit signal is directly converted from digital to analog (D / A), there has been an opinion that the nuance is unsatisfactory in the delicate part of the reproduced sound quality.

  On the other hand, in D / A conversion of acoustic signals, for example, 24-bit conversion means have become available at low cost, and it is conceivable to improve sound quality by using such conversion means. However, for example, if an expansion is performed by simply inserting an 8-bit value 0 below 16-bit information, a large amount of noise is generated and the sound quality is significantly degraded. Moreover, if such noise is removed using a low-pass filter, the sound will be felt as a whole.

The present invention has been made in view of such problems, and an object of the present invention is to extend the number of bits of an information signal quantized with a small number of bits, for example, recording on a compact disk. At the same time, it prevents the deterioration of information such as the occurrence of noise during the expansion. Thus, it is possible to provide a directly applicable signal processing equipment to record, such as a conventional compact disc.

For purposes of resolution to the present invention the above problems, the present invention is a signal processing device for extending the number of bits of quantized original signals, the co Extending the number of bits of quantized original signals The filter means for cutting off the harmonic component due to the extended bits and the first calculation means for subtracting the output of the filter means from the signal obtained by adding an arbitrary lower bit for the extension to the original signal . Then, the peak moving correction average value is calculated with a time constant longer than the quantization sampling with respect to the output of the first calculating means, and the amplitude of the output of the first calculating means is calculated by the calculated moving correction average value. And a second arithmetic unit that limits and multiplies the output of the first arithmetic unit. And a third calculating means for adding the output of the second calculating means to the output of the filter means .

In the signal processing apparatus according to the preferred embodiment of the present invention, the filter means is formed of a variable filter including delay means and a correction unit for controlling the coefficient thereof, and the correction unit includes the quantized original signal and the original signal. And a fourth computing means for calculating an error ε between the output of the low-pass filter whose cutoff frequency is controlled in accordance with the external signal and the output of the variable filter . The coefficient of the variable filter is controlled using a correction algorithm that minimizes the error ε, which is the output of the operation means 4 .

Further, in the signal processing device of the preferred embodiment of the present invention, the original signal is a signal obtained by quantizing the acoustic signal, and the external signal determined by the tendency of the original signal is different from the conversation of the acoustic signal and the music, and / Or a signal indicating the genre of music.

According to the present invention, for example, the number of bits of an information signal quantized with a small number of bits as in recording on a compact disc is expanded, and information deterioration such as generation of noise does not occur during the expansion. it can be, thereby, it is possible to provide a directly applicable signal processing equipment to record, such as a conventional compact disc.

Hereinafter, with reference to the drawings in describing the present invention, FIG. 1 is a block diagram showing a configuration of an embodiment of an audio signal processing apparatus according to the signal processing equipment according to the present invention.

  In FIG. 1, for example, a 16-bit PCM signal is supplied to the input terminal 1. A signal from the input terminal 1 is supplied to the variable filter 3 through a predetermined delay means 2. A signal from the input terminal 1 is supplied to the arithmetic means 5 through the low pass filter (LPF) 4, and a difference from the output of the variable filter 3 is calculated and supplied to the coefficient correcting means 6. The coefficient obtained by the coefficient correcting means 6 is supplied to the variable filter 3.

  Further, the signal from the input terminal 1 is supplied to the lower bit adding circuit 8 through a predetermined delay means 7, and for example, 8 bits of value 0 are inserted into the lower order of the 16-bit input signal to obtain 24 bits. This 24-bit signal is supplied to the calculation means 9 and the difference from the output of the variable filter 3 is calculated. Then, this difference value (24 bits) is added to the output of the variable filter 3 by the addition circuit 11 through the weighting circuit 10 and taken out to the output terminal 12.

  Also, the difference value from the calculating means 9 is supplied to the peak detection circuit 13, and the peak detected signal is supplied to the limiter circuit 15 through the low pass filter (LPF) 14, and the signal corresponding to the lower 8 bits is deleted. The weighting circuit 10 performs weighting based on the output of the limiter circuit 15. As a result, the number of bits of the information signal is expanded, and correction processing for information deterioration due to the expansion is performed.

  Further, in this circuit, the variable filter 3 is, for example, an FIR type digital filter as shown in FIG. 2, and an input signal is supplied to a plurality of stages of unit delay means Z-1 connected in cascade. The output is added by the adding means (+) through the weighting circuit h. The added value is taken out as an output value y (n), the difference from the target value d (n) is calculated by the calculating means 5, and the weight correcting circuit h is used by the coefficient correcting means 6 so that the difference value is minimized. The coefficient is obtained.

That is, if the signal supplied to the variable filter 3 is x (n), the output value y (n) is expressed as follows.

The difference value (error signal) ε (n) is ε (n) = d (n) −y (n).
And the mean square error e
e = E {ε ² (n)}
The coefficient is obtained by the coefficient correction means 6 so that the value is minimized.

Therefore, the condition for minimizing the mean square error e is:
^{e = E {d 2 (n} )} - 2E {d (n) · y (n)} + E {y 2 (n)}
As
E {d ² (n)} = Pd
Then, it becomes as follows.

であるから、

である。 That is,

Because

It is.

on the other hand,
x (n) = Sin (2πn / N)
And
d (n) = Cos (2πn / N) + s (n)
Where s (n) is a white signal,
s (n) = E {s ² (n)} = σ 2
It is.

  Here, for example, when 8 bits of value 0 are inserted in the lower order in the circuit shown in FIG. 3B with respect to a sine wave signal having a frequency of 1 kHz as shown in FIG. The signal quantized as shown in (a) of C changes so that the level difference of each quantization step becomes large as shown in (b) of FIG. The frequency spectrum of such a signal is a single sinusoidal wave signal as shown in FIG. 3D (a), but by inserting bits, the frequency spectrum becomes higher as shown in FIG. Waves are generated.

  On the other hand, by providing a low-pass filter (LPF) as shown in FIG. 4A, it is shown in FIG. 4B from a signal in which harmonics are generated as shown in FIG. Thus, harmonic components can be removed. However, this alone will result in a muffled sound. Therefore, as shown in FIG. 5A, a signal obtained by deleting the original quantization and lower bits from the difference signal between the input signal and the output of the low-pass filter (LPF) is output to the low-pass filter (LPF). to add.

  According to this, the signal quantized as shown in (a) of FIG. 5B is smoothed by the low pass filter (LPF) as shown in (b) of FIG. As shown in (c), by adding the high frequency component deleted by the low pass filter (LPF), it is possible to prevent a muffled sound. That is, in this case, the band of the low pass filter (LPF) has a high sound quality of 24 bits and the high sound is directly added through and is prevented from becoming a muffled sound.

  4 and 5 uses a low-pass filter (LPF) as means for smoothing the waveform. However, this has a problem that the cutoff frequency is fixed. Therefore, an adaptive filter (LMS) using delay means is used as shown in FIG. That is, such an adaptive filter is equivalent to the variable filter shown in FIG. 1 and FIG. 2, and thereby, from a signal including harmonic components as shown in FIG. 6B (a). As shown in FIG. 5B, the harmonic components can be adaptively removed to smooth the waveform.

  In the circuit of FIG. 6A, attention is paid to a continuous signal and it is extracted by an adaptive filter (LMS). However, even in a continuous signal, there are many quantization noises at high frequencies. To do. Therefore, as shown in FIG. 7, the low-pass filter (LPF) is used to remove high-frequency quantization noise and correct the coefficient of the adaptive filter (LMS). The cutoff frequency of the low-pass filter (LPF) is generally about 10 kHz when the sampling frequency is 44.1 kHz, that is, half the frequency of Nyquist.

  Furthermore, the cut-off frequency of the low-pass filter (LPF) can be changed, for example, by narrowing the pass band, depending on the conversation of the original acoustic signal and the music, the music genre, and the like. Thereby, for example, better sound quality improvement can be performed according to the genre of music.

  That is, in FIG. 1 described above, for example, music information of a compact disc (CD) is 16 bits. Therefore, an adaptive filter is formed by the delay unit + variable filter unit, and the variable filter control is performed by the least square method so as to approach the signal source expanded to 24 bits by removing the harmonics from the reference signal through the LPF. As a result, the continuous (highly reproducible signal) due to the delay is interpolated from 16 bits to 24 bits with a filtering effect. However, since this sound is a reproducible signal, it becomes a muffled sound.

  Therefore, next, the lower 8 bits of the source signal delayed by the variable filter processing are added (the lower 8 bits are 0), and the difference from the adaptive filter is calculated. This differential signal is a high-frequency signal, but in the case of a signal with a magnitude less than the quantization error, it can be expanded to 24 bits by not adding it, and the conventional signal is added to the high-frequency information. So you can solve the problem.

  In addition, in the peak detection of FIG. 1, when it is determined whether or not to add based on the magnitude of the quantization error, there is a possibility that a pop noise may be generated since a scene of adding or not abruptly occurs. . Therefore, it is possible to remove the pop noise because the high frequency signal can be smoothly added by calculating the quantization error as 1.0 at the maximum in peak detection and multiplying the error amount according to the magnitude of the error. it can.

Furthermore, Oite the signal processing equipment of the present invention, by providing a genre of LPF (or fixed LPF), only the low-frequency portion of the target signal, can be targeted value of the adaptive filter. According to this, it is possible to extend the bit precision to 24 bits by the LPF, and the adaptive filter does not follow the high-frequency signal even if the signal has high continuity.

  In the case of a high-frequency signal, there is a cyclic signal due to a quantization error, so that it does not react to such a signal. In addition, by detecting the peak of the error signal and using the moving correction average (LPF), it is possible to suppress an excessive response to a sudden music signal. As a result, pop noise can be eliminated. Further, by limiting the peak detection and controlling the amplitude of the error signal, the rate of change can be made square.

Thus, according to the signal processing equipment of the present invention, for example, a sound source of a compact disk (CD) is a 16-bit, not enough when viewed in a plane referred to as high-quality sound. On the other hand, a 24-bit DAC can be used at low cost, and there are many 24-bit sound sources. Therefore, the present invention has an effect of improving sound quality by predicting and extending the lower 8 bits, and can easily cope with a conventional CD.

  In FIG. 8, the effect of the present invention will be described with reference to waveforms. Here, A in FIG. 8 shows a waveform immediately after the start of calculation, and here, quantization noise is generated. On the other hand, B in the figure shows a waveform after 100 ms from the start of the calculation, which indicates that the quantization noise is reduced.

Therefore, in the above embodiment, the number of bits of the information signal quantized with a small number of bits, for example, recording on a compact disc, is expanded, and information deterioration such as generation of noise occurs during the expansion. it can ensure that no Thereby, it is possible to provide a directly applicable signal processing equipment to record, such as a conventional compact disc.

  Thus, according to the signal processing device of the present invention, the signal processing device expands the number of bits of the quantized signal, and expands the number of bits of the quantized original signal and generates harmonic components by the expanded bits. Filter means for cutting off the output, first calculation means for subtracting the output of the filter means from a signal obtained by adding an arbitrary lower bit of the extension to the original signal, and the output of the first calculation means as the output of the filter means By having the second calculation means for adding to, a good number of bits can be expanded.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the claims.

It is a block diagram showing a configuration of an embodiment of an audio signal processing apparatus according to the signal processing equipment according to the present invention. It is a block diagram which shows the structure of the principal part. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a wave form diagram for explanation of the effect.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Input terminal, 2 ... Delay means, 3 ... Variable filter, 4 ... Low pass filter (LPF), 5 ... Calculation means, 6 ... Coefficient correction means, 7 ... Delay means, 8 ... Lower bit addition circuit, 9 ... Calculation means DESCRIPTION OF SYMBOLS 10 ... Weight circuit, 11 ... Adder circuit, 12 ... Output terminal, 13 ... Peak detection circuit, 14 ... Low pass filter (LPF), 15 ... Limiter circuit

Claims (4)

A signal processing device for extending the number of bits of a quantized original signal,
Co Extending the number of bits of quantized original signals, filter means for blocking the higher harmonic component due to extended bits,
The output of said filter means, first calculating means for subtracting from the added any low-order bits of the extended component in respect original signal signal,
A peak shift correction average value is calculated with a time constant longer than the quantization sampling with respect to the output of the first calculation means, and the output amplitude of the first calculation means is calculated based on the calculated shift correction average value. And a second computing means for multiplying the output of the first computing means by limiting
And a third arithmetic means for adding the output of the second arithmetic means to the output of the filter means . The filter means is formed of a variable filter including delay means and a correction unit for controlling the coefficient thereof,
The correction unit receives an input of the quantized original signal and an external signal determined by the tendency of the original signal, an output of a low-pass filter whose cutoff frequency is controlled according to the external signal, and a variable filter Including a fourth calculation means for calculating an error ε with respect to the output, and controlling a coefficient of the variable filter using a correction algorithm that minimizes the error ε, which is an output of the fourth calculation means. The signal processing apparatus according to claim 1.   The signal processing apparatus according to claim 1, wherein the original signal is a signal obtained by quantizing an acoustic signal.   The signal processing apparatus according to claim 2 or 3, wherein the external signal determined by the tendency of the original signal is a signal indicating a conversation and music of the acoustic signal and / or a genre of music.

Images