KR19990056431A - Bit Allocation Method in Digital Audio Coding Device - Google Patents

Abstract

The present invention provides a method for allocating bits in a digital audio encoding apparatus by determining bit allocation order for each subband according to a human perceptual importance according to a frequency band. In the bit allocation method according to the present invention, a first mask-to-noise ratio (MNR) is calculated for each subband by applying an auditory psychological model of an audio signal divided into subbands, and a first mask. The second mask-to-noise ratio is calculated by applying a weight set according to the number of critical bands present in each subband to the noise-to-noise ratio, the minimum value of the calculated second mask-to-noise ratio is detected, and the sub corresponding to the detected minimum value. And allocating bits to the band, and sequentially repeating the above steps until the remaining bits that can be allocated to the audio signal at the corresponding time after the bit allocation remain as the minimum bits that can be allocated to one subband. do. Therefore, the sound quality can be prevented from being degraded even at a low data rate.

Description

Bit Allocation Method in Digital Audio Coding Device

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bit allocation method in a digital audio encoding apparatus, and more particularly, to a bit allocation method for allocating bits by determining the bit allocation order according to the perceptual importance of each subband.

The digital audio coding apparatus is proposed to transmit a high quality audio signal with a small data transmission bit rate, and is mainly based on a subband coding scheme. That is, the input audio signal is divided into 32 subbands, the scale factor for each divided subband is determined, and the result of fast Fourier transform of the simultaneously input audio signal and the calculated scale The psychoacoustic model for each subband by the factor is obtained. As shown in FIG. 1, the allocation bit for each subband is determined by applying the psychoacoustic model for each subband.

That is, when the psychoacoustic model for each subband is transmitted, each subband is obtained by using a signal to mask ratio (SMR) value obtained from the auditory psychological model and a signal to noise ratio (SNR) value that varies according to bit allocation in step 101. The mask-to-noise ratio (MNR) value for each band is obtained by Equation 1.

MNR = SNR-SMR ｜ _dB

In operation 102, the subband having the minimum MNR value is detected. If a subband having the minimum MNR value is detected, the flow proceeds to step 103 and a predetermined bit is allocated to the detected subband. At this time, the predetermined bit is a bit allocation unit per subband. Accordingly, when one bit is used in step 103 when the used bit is 36 bits (in the case of having 36 sample structures per subband), 36 bits are substantially allocated to the corresponding subband. Therefore, when the maximum bit that can be allocated to one audio signal is 1500 bits, if the first 1 bit is allocated in step 103, the 36 bits are subtracted to 1464 bits. Each time one bit is allocated, the remaining bits that can be allocated by subtracting 36 bits are obtained.

When the bits are allocated to the subband having the minimum MNR and the remaining assignable bits are calculated accordingly, the flow proceeds to step 104 and the remaining number of bits that can be allocated (the remaining bits) and the minimum bits that can be allocated to one subband. Compare the number MIN. Here, the minimum number of bits MIN is set to 36 when each subband has a 36 sampling structure as described above. As a result of the comparison of step 104, when the remaining number of bits is larger than the minimum number of bits MIN, the process returns to step 101 and repeats the above-described process. However, as a result of the comparison, if the remaining number of bits is not greater than the minimum number of bits MIN, bit allocation for the audio at that time is terminated. As described above, bits for each subband are allocated without considering the perceptual importance of each human frequency band.

Human hearing organs, however, have different degrees of importance when recognizing sound. That is, in the case of the low frequency sound as shown in FIG. 2, the threshold band according to the frequency has a fine characteristic, whereas in the case of the high frequency sound, the change of the threshold band according to the frequency tends to be insensitive. Therefore, although a large number of bits should be allocated first in a region where the change of the critical band is sensitive, the low frequency band is more psychologically important by determining the order of subbands to allocate bits by the detected MNR using SNR and SMR. There is a phenomenon in which fewer bits are allocated. This phenomenon occurs more seriously when operating at lower data rates.

Disclosure of Invention The present invention has been made to solve the above-described problem, and an object thereof is to provide a bit allocation method for allocating bits by determining the order of subbands to which bits are allocated in consideration of the perceptual importance of each frequency band. have.

In order to achieve the above object, in the digital audio encoding apparatus according to the present invention, the bit allocation method is applied to the psychoacoustic model of an applied audio signal to allocate bits for each subband. Calculating a first mask-to-noise ratio (MNR) by subtracting a signal-to-noise ratio (SNR) that varies according to bit allocation from the signal-to-mask ratio (SMR) obtained from the method; The weight set by the number of critical bands for each subband to the calculated first mask-to-noise ratio W _sb ) To apply a second mask to noise ratio for each subband ( Weighted Calculating MNR); The second mask-to-noise ratio for each subband calculated in the calculation step ( Weighted Retrieving a minimum value of MNR); If a minimum value is found, allocating a predetermined bit to a subband corresponding to the found minimum value; Allocating bits in the bit allocation step and returning to the first mask-to-noise ratio calculation step if the remaining number of bits that can be allocated to the currently applied audio signal is larger than the minimum number of bits that can be allocated to one subband; If the remaining number of bits is not greater than the minimum number of bits, the step of terminating the bit allocation of the currently applied audio signal, characterized in that performed.

1 is a flowchart illustrating a conventional bit allocation method in a digital audio encoding apparatus.

2 is a diagram of human auditory characteristics with respect to frequency,

3 is a functional block diagram of a general digital audio encoding apparatus to which the method according to the present invention is applied,

4 is a flowchart of a bit allocation method according to the present invention;

5 is a characteristic diagram of weight coefficients WF for each subband.

<Description of the symbols for the main parts of the drawings>

301: subband split filter 302: scale factor calculator

303: Fast Fourier Transform 304: Psychological psychology modeling unit

305: bit allocation unit 306: quantizer

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 3 is a block diagram of a general digital audio encoding apparatus to which the method according to the present invention is applied. When an audio signal is applied, FIG. 3 shows a subband segmentation filter 301 and a subband segmentation filter 301 which are divided into a predetermined subband and transmitted. A scale factor calculation unit 302 for calculating a fixed scale factor for an audio signal divided and transmitted for each frequency band, a FFT 303 for fast Fourier transforming an input signal, and a conversion result and scale output from the FFT 303 The psychoacoustic modeling unit 304 outputs the psychoacoustic model for each subband using the scale factor for each subband provided by the factor calculating unit 302, and the psychoacoustic model output from the psychoacoustic modeling unit 304. The bit allocation unit 305 for allocating bits per subband is divided and transmitted from the subband division filter 301 according to the bits allocated from the bit allocation unit 305. It consists of a quantizer 306 for quantizing the sub-band-specific processing.

4 is a flowchart illustrating a bit allocation method according to the present invention.

Next, an operation of the method illustrated in FIG. 4 will be described with reference to FIG. 3.

First, auditory psychology on an audio signal subband-divided in the conventional manner by the operation of the subband split filter 301, the scale factor calculator 302, the FFT 303, and the auditory psychological modeling unit 304 is performed. When the model is transmitted, the bit allocator 305 first calculates a mask to noise ratio (MNR) for each subband by Equation 1 in step 401. In operation 402, a weighted MNR obtained by assigning a weight to each subband to the calculated MNR is calculated as in Equation 2.

Weighted MNR = W _sb * MNR

At this time, the weight for each subband is the number of critical bands included for each subband. The critical band exists for each frequency band as shown in the example shown in FIGS. 5A and 5B. That is, according to the example shown in Figs. 5A and 5B, there are five threshold bands from 0 to 500Hz frequency band, three threshold bands from 501Hz to 1000Hz, and two threshold bands from 1001Hz to 1500Hz. do. As such, the lower frequency band has more threshold bands. The weight for each subband is set by the number of such critical bands. That is, the weight of one subband (0 to 500Hz) is 5, the weight of the 2 subbands (501 to 1000Hz) is 3, and the weight of the 3 subbands (1001 to 1500Hz) is 2. The weight set in this way ( W _sb ) And MNR obtained by Equation 1 are calculated as Equation 2 to calculate the weighted MNR.

The process proceeds to step 403 to detect a subband having the minimum MNR among the weighted MNRs, and proceeds to step 404. In step 404, a predetermined bit is preferentially allocated to the subband detected in step 403. In this case, the allocated bits are allocated bits per subband, and the bits allocated substantially are determined according to the number of samples per subband. That is, when 36 samples exist for each subband, when 1 bit is allocated in step 404, the bits allocated to the corresponding subbands are 36 bits.

If a predetermined bit is allocated to the corresponding subband, the flow proceeds to step 405 where the bit is allocated and the number of remaining bits is compared with the minimum number of bits. Here, the minimum number of bits is the minimum bit that can be allocated to one subband, and is set to 36 when the subband has 36 samples as described above.

As a result of the comparison in step 405, if the number of remaining bits is greater than the minimum number of bits, the process returns to step 401 to repeat the above-described process. However, if the remaining number of bits is not greater than the minimum number of bits, bit allocation for the currently applied audio signal is terminated.

In the case of such a bit allocation, when the subband to which 16 bits (16x36 bits) are allocated is generated and returned from step 405 to step 401, the MNR for the subband to which 16 bits are allocated is maximum. Not calculated.

As described above, the present invention uses an MNR value in consideration of a weight set by the number of critical bands present for each subband when allocating bits for each subband by applying an psychoacoustic model in encoding a digital audio signal. By determining the order of subbands to which bits are allocated, there is an advantage in that sound quality is prevented from being degraded when transmitting audio signals at low data rates by always assigning bits to subbands in a low frequency band.

Claims (2)

A digital audio encoding apparatus for allocating bits for each subband by applying to an auditory psychological model of an applied audio signal, Calculating a first mask-to-noise ratio (MNR) by subtracting a signal-to-noise ratio (SNR) variable according to bit allocation from the signal-to-mask ratio (SMR) obtained from the psychoacoustic model; The weight set by the number of critical bands for each subband in the calculated first mask-to-noise ratio W _sb ) To apply a second mask to noise ratio for each subband ( Weighted Calculating MNR); The second mask-to-noise ratio for each subband calculated in the calculating step ( Weighted Retrieving a minimum value of MNR); If the minimum value is found, allocating a predetermined bit to a subband corresponding to the found minimum value; After allocating bits in the bit allocation step, if the remaining number of bits that can be allocated to the currently applied audio signal is larger than the minimum number of bits that can be allocated to one subband, the method returns to the first mask-to-noise ratio calculating step. step; And if the remaining number of bits is not greater than the minimum number of bits, terminating bit allocation of the currently applied audio signal. The method of claim 1, wherein the calculating of the second mask-to-noise ratio multiplies the weight by the first mask-to-noise ratio (MNR). Weighted MNR) to obtain a bit allocation method.