JP2952878B2 - Digital signal processor - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital signal processing device, and is preferably applied to, for example, a digital signal processing device configured to record, reproduce, and transmit an audio signal or the like with high quality. Things.

B. Summary of the Invention The present invention can improve the signal-to-quantization noise ratio in the digital signal processing device by setting the order of the noise filter higher than that of the prediction filter.

C Conventional technology Conventionally, in this kind of digital signal processing device,
Adaptive predictive coding (APC)
By encoding the audio signal and compressing the information using the method described above, there is a method that prevents deterioration of the S / N ratio, clarity, etc. beforehand and transmits the audio signal with high transmission efficiency ( JP-A-59-223033, JP-A-60-23033
223034, JP-A-61-158217, JP-A-61-15
No. 8218).

That is, in FIG. 4, 1 denotes a digital signal processor, provides the input digital signal S _I to the prediction of filter 3.

As shown in FIG. 5, the predicting filter 3 weights the output signals of the two delay circuits 4A and 4B connected in series by the multipliers 5A and 5B and then adds the output signals by the adder 6. Of the filter circuit.
And the filter characteristic is determined by the weighting amount of 5B.

In digital signal processing apparatus 1, by providing the output signal of the prediction of the filter 3 together with the input digital signal S _I to the adder 7 and the difference signal of the output signal and the input digital signal S _I of the prediction of the filter 3 residues Obtain the difference signal S _Z1 .

Linear prediction analyzer 8 receives the residual signal S _Z1, thereby inputting the spectrum shape of the digital signal S _I is detected at predetermined time intervals, prediction of outputs a parameter signal S _P based on the detection result The filter characteristics of the filter 3 are switched.

That is, the spectrum of the input digital signal S _I is,
Multipliers 5A and 5B if distributed in high frequency band
The amount of weighting is selected to a value 0 (hereinafter referred to as straight PCM mode), to match the spectrum shape of the input digital signal S _I spectrum shape of the residual signal S _Z1.

On the other hand, when the spectrum gradually approaches the flat spectrum shape from the straight PCM mode, the weighting amounts of the multipliers 5A and 5B are changed to values 0.9375 and 0 according to the spectrum shape.
(Hereinafter referred to as first-order differential PCM mode), the values 1.968875 and-
Switch to 0.8125 (hereinafter called secondary differential PCM mode).

Thereby, in the prediction filter 3, the residual signal S
_The filter characteristics are switched so that _Z1 becomes smaller.

Further linear predictive analyzer 8 outputs the parameter signal S _P to the prediction of the filter 9 and transmission predictive of filter 10 of the subject, multipliers floating coefficient signal S _F based on the maximum value of the residual signal S _Z1 outputs 11, thereby inputting the corrected residual signal S _Z1 to the quantizer 12 to a predetermined dynamic range.

That is, the requantizer 12 receives the residual signal _SZ1 via the adder 13 and the multiplier 11, requantizes the residual signal _SZ1 , and sends out the residual signal _SZ1 to the transmission target.

On the other hand, on the transmission target side, the transmission signal S _L1 transmitted to the transmission path L1 is
The prediction signal is received by the prediction filter 10 via the adder 20 and the adder 20, and the output signal of the prediction filter 10 is fed back to the adder 20.

Thus by switching the same filter characteristics as predictive of filter 3 based on the prediction of the filter 10 to the parameter signal S _P, obtained by decoding the transmission signal S _L1, thereby the input digital signal S _I residual signal instead of S _Z1 min transmitting the, is configured so as to be able to transmit the input digital signal S _I at a high transmission efficiency.

In this case the quantizer 12 via an adder 21 to obtain a difference signal S _Z2 input and output signals, a multiplier 22 and the prediction of the filter 3 becomes the difference signal S _Z2 in inverse characteristic of the multiplier 11 the By feeding back to the adder 13 via the predictive filter 9 having characteristics, quantization noise generated at the time of requantization (that is, a requantization error signal and hereinafter referred to as requantization noise) is suppressed. Has been made.

By the way, in this kind of digital signal processing apparatus, by changing the spectrum shape of the requantization noise by using a noise shaping method, the signal-to-quantization noise ratio (SNR) on perception is improved. What was done is proposed (IEEE TRANSACTIONS ON ACOUSTIC
S, SPEECH, AND SIGNAL PROCESSING, VOL.ASSP-27, NO.3, J
UNE 1979, IEICE Journal 4 / '87 VOL.70, NO.4 page 3
92-400, JP-A-59-223032, JP-A-60-103746
JP-A-61-158220).

This noise shaping method is to improve the signal-to-quantization noise ratio by using the masking effect on hearing by approximating the spectrum shape of the requantization noise to the spectrum shape of the audio signal. I have.

That is, in the configuration of FIG. 4, for example, in the straight PCM mode, the weighting amount of the prediction filter 9 is selected to be different from the weighting amount of the prediction filter 3 so that the spectrum shape of the requantization noise is changed to the audio signal. Can be approximated to the spectrum shape of
As a result, the signal-to-quantization noise ratio can be improved.

Problems to be Solved by the Invention D However, when the noise shaping method is actually used in the configuration of FIG. 4, there is a problem that the effect of improving the signal-to-quantization noise ratio is still insufficient.

The present invention has been made in consideration of the above points, and has as its object to propose a digital signal processing device capable of improving the signal-to-quantization noise ratio as compared with the related art.

Means for Solving Problem E In order to solve such a problem, in the present invention, a prediction filter for calculating a predicted value of an input signal and a prediction for outputting a difference signal between the input signal and the predicted value of the prediction filter are provided. Error detection means, requantization means for requantizing and outputting the difference signal, F (z) representing a spectrum characteristic suitable for auditory characteristics, and F (z) representing a filter characteristic R (z), and a filter characteristic P (z) of a prediction filter And a noise filter determined based on the following equation: R (z) = F (z) + P (z) -F (z) · P (z), and a requantization error generated at the time of requantization. The signal is fed back to the requantization means via a noise filter.

F function Based on the filter characteristic F (z) using the filter characteristic R (z) of the noise filter, F (z) representing the characteristic of the noise spectrum suitable for the auditory characteristic, and the filter characteristic P (z) of the prediction filter. With this configuration, the spectrum shape of the requantization noise can be made suitable for the auditory characteristics.

G Embodiment In FIG. 1, in which parts corresponding to those in FIG. 4 are assigned the same reference numerals, reference numeral 40 designates a digital signal processing apparatus having a noise shaping function as a whole,
, A noise filter 41 is provided.

As shown in FIG. 2, the noise filter 41 includes n stages of series-connected delay circuits 42A, 42B,.
Multiplier 43 for weighting the output signals of 2A, 42B,..., 42X
, 43X and adders 44B,..., 44X for adding the weighted output signals.

Therefore, by using a high-order noise filter 41 instead of the predictive filter 9 of the same order as the conventional predictive filter 3, the requantized noise has a spectrum closer to a desired shape than the conventional one. It can be shaped into a shape.

That is, as shown in FIG. 3, in the straight PCM mode, the first-order differential PCM mode, and the second-order differential PCM mode, the noise spectrum shape S of the requantized noise included in the output signal S ₀ demodulated on the transmission target side. For example, _n is represented by, for example, the following formula: S _n = 1−1.33678 Z ⁻¹ +0.64 Z ⁻² (1) S _n = 1−0.5 Z ⁻¹ (2) S _n = 1−0.3 Z ⁻¹ If the spectrum shape is adapted to the auditory characteristics represented by (3), the signal-to-quantum is determined by the perceptual curve of the loudness of the pure tone and the auditory characteristics on the frequency axis indicated by the masking effect on the frequency axis. It is clear from the result of the simulation experiment that the ratio of the noise to be improved is improved.

When selecting a spectrum shape S _n of the desired re-quantization noise, where the following equation ^{F (z) = 1.33678Z -1 +} 0.64Z -2 ...... (4) F (z) = 0.5Z -1 ...... (5) F (z) = 0.3Z -1 ...... at (6), (4) - (6) are collectively equation equation _{S n = Δ (1-F} (z)) ...... (7) Can be represented by

Thus, in the primary differential PCM mode and the secondary differential PCM mode spectrum of successively higher frequency band from the straight PCM mode is decreased, the spectrum shape of the requantization noise S _n according to the decrease of the spectrum, The spectrum can be sequentially shaped into a flat spectrum shape, and the signal-to-quantization noise ratio can be improved by using the auditory masking effect even in modes other than the straight PCM mode.

By the way, the frequency characteristic of the prediction filter 3 is represented by P (Z),
Placing the frequency characteristic of the noise filter 41 and R (Z), the spectrum shape S _'n of the actual re-quantization noise by the prediction of filter 3 and the noise filter 41 at the Δ a flat frequency characteristic, the following equation It can be expressed as

Here, the spectrum shape S _'n requantization noise by the prediction of filter 3 and the noise filter 41, if matching the spectrum shape S _n of the requantization noise suitable for the desired auditory characteristic, (7) And from equation (8), Is obtained.

 Therefore, by solving the equation (9), the following equation is obtained: R (Z) = F (Z) + P (Z) -F (Z) · P (Z) (10)

Here, the filter characteristics of the actual prediction filter 3 in the straight PCM mode, the first-order difference PCM mode, and the second-order difference PCM mode are as follows.
Since the values are set to 0.9375 and 0, and the values are 1.796875 and -0.8125, P (Z) is represented by the following equation: P (Z) = 0 (11) P (Z) = 0.9375Z ^-1. (12) P (Z) = 1.796875Z ⁻¹ −0.8125Z ⁻² (13)

Therefore, by substituting equations (5) to (7) and equations (11) to (13) into equation (10), the following equation is obtained: R (Z) = 1.33678Z ^-1 + 0.64Z ^-2 (14) R (Z) = 1.4375Z ^-1 + 0.46875Z ^-2 ... (15) R (Z) = 2.06875Z ^-1 -1.3515632Z ^-2 + 0.24375Z ^-3 ... (16) This makes it possible to set the filter characteristic R (z) of the noise filter 41 so that the noise spectrum of the output on the transmission target side has a filter characteristic that can aurally improve the signal-to-quantization noise ratio. . In this case, in this example, 1 is used in the straight PCM mode.
Value of the weighting coefficient of the multipliers 43A and 43B in the second and third stages
With 1.33678 and 0.64, it can be seen that the third and subsequent stages should have a value of 0.

Further, in the first-order differential PCM mode, the weighting coefficients of the first and second stage multipliers 43A and 43B may be set to values of 1.4375 and 0.46875, and the third and subsequent stages may be set to a value of 0.
In the next difference PCM mode, the weighting coefficients of the first-stage, second-stage, and third-stage multipliers 43A, 43B, and 43C are set to a value of 2.09687.
It is understood that it is only necessary to set 5, 5, -1.351563, and 0.24375 to the value 0 in the fourth and subsequent stages.

Thus, in the straight PCM mode, the first-order differential PCM mode, and the second-order differential PCM mode, by switching the weighting coefficient and switching the filter characteristic of the noise filter 41, the spectrum shape of the requantized noise is changed from (1) to
It can be shaped into the shape represented by equation (3).

According to the above configuration, the order of the noise filter 41 is set higher than that of the prediction filter 3, and the filter characteristic of the noise filter 41 is switched in accordance with the filter characteristic of the prediction filter 3, thereby achieving requantization. The spectrum shape of the quantization noise can be shaped into a desired shape, and thus the signal-to-quantization noise ratio can be improved as compared with the related art.

In the above-described embodiment, the case where the present invention is applied to the digital signal processing device having the straight PCM mode, the first differential PCM mode, and the second differential PCM mode has been described. However, the present invention is not limited to this. The present invention can be widely applied to a digital signal processing device configured to transmit a digital signal using the adaptive prediction coding method.

H Effects of the Invention As described above, according to the present invention, the filter characteristic R (z) of the noise filter is represented by F (z) representing the spectrum characteristic suitable for the auditory characteristic and the filter characteristic P of the prediction filter.
By determining based on the expression of the filter characteristic R (z) using (z), the spectrum shape of the requantization noise can be made suitable for the auditory characteristic, and thus the requantized noise is requantized. The output signal-to-quantization noise ratio can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a digital signal processing apparatus according to the present invention, FIG. 2 is a block diagram showing a noise filter, FIG. 3 is a characteristic curve diagram showing a spectrum shape of requantization noise, and FIG. FIG. 1 is a block diagram showing a conventional digital signal processing apparatus, and FIG. 5 is a block diagram showing its prediction filter. 1, 40 ... a digital signal processing device, 3, 9, 10 ... a prediction filter, 12 ... a requantizer, 41 ... a noise filter.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-61-158217 (JP, A) JP-A-61-158218 (JP, A) JP-A-61-158219 (JP, A) JP-A 61-158219 158220 (JP, A) JP-A-62-3516 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) H03M 7/38

Claims (1)

(57) [Claims] 1. A prediction filter for calculating a prediction value of an input signal, prediction error detection means for outputting a difference signal between the input signal and the prediction value of the prediction filter, and requantization of the difference signal R (z) using a re-quantization means for outputting a filter characteristic R (z) and a filter characteristic F (z) representing a spectrum characteristic suitable for an auditory characteristic and a filter characteristic P (z) of the predictive filter. = F (z) + P (z) -F (z) · P (z), and a re-quantization error signal generated at the time of the re-quantization is passed through the noise filter. A digital signal processing device, wherein the digital signal is returned to a requantization unit.