JP2001109497A - Audio signal encoding device and audio signal encoding method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an audio signal encoding device and an audio signal encoding method by which the number of bits required for encoding is reduced and the deterioration of sound quality is prevented without reducing the number of quantized bits for encoding even if the number of bits lacks as a whole. SOLUTION: A hearing model calculation part 3 suppresses the number of bits allocated to frequency spectral data so that the number is decreased without dropping a masking level showing a hearing model even if the part 3 selects a second conversion length whose frequency resolution is low at the time of calculating the hearing model.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an audio signal encoding apparatus and an audio signal encoding method for compressing and encoding a digitized audio signal by utilizing human auditory characteristics.

[0002]

2. Description of the Related Art In recent years, digitalization of audio devices has been rapidly progressing. For example, ATRAC (A) of an audio signal encoding method adopted for a mini disc (MD)
adaptive Transform Acoustic
c Coding) system and various types of MPEG, which is an audio signal encoding system adopted in digital satellite broadcasting, are intended to efficiently compress the original signal while maintaining the sound quality of the reproduced signal when decoding and reproducing. Digital audio signal processing technology.

As one of audio signal encoding techniques for compressing these digital audio signals, an input digital audio signal is converted to a QMF (Quadrature M) signal.
(e.g., color filter) to divide the data into time-series sub-band data for each of a plurality of frequency components, and allocate the number of quantization bits to each sub-band data by utilizing human auditory characteristics to perform quantization. Subband coding (Subband Co
ding) method.

[0004] Separately, an input digital audio signal is converted to a fast Fourier transform (FF) as an input signal.
T: Fast Fourier Transform)
And discrete cosine transform (DCT: Discrete Co
sine Transform, Modified Discrete Cosine Transform (MDCT)
A time-frequency conversion process such as Cosine Transform is performed to convert the frequency spectrum data into a plurality of frequency spectrum data. Perform Transform Cod
ing) method.

[0005] In the MPEG AAC method, MDCT is used to convert a time-series digital audio signal into frequency spectrum data, which is one of the conversion coding methods. Further, there is an encoding system in which a subband encoding system and a transform encoding system are combined, and ATRAC of MD (mini disc) is one of the encoding systems. In any of the above encoding schemes for compressing digital audio signals, the original signal is compressed by using human auditory characteristics. Human auditory characteristics include a minimum audible characteristic that indicates the minimum perceptible signal level that differs for each frequency in a quiet state.

[0006] Further, between signals of different frequencies generated simultaneously, a sound having a high signal level makes it difficult to hear a sound having a low signal level. There is a time masking effect that makes it difficult to hear a loud sound with a low signal level. Using these human auditory characteristics, thinning out signals that are not perceived or reducing the number of quantization bits of signals that are hardly perceived achieves a high compression rate while maintaining sound quality during reproduction.

FIG. 4 is a block diagram showing a configuration of a conventional audio signal encoding apparatus which realizes encoding processing of the transform encoding method. In FIG. 4, 41 is the number M in time series.
Is a first conversion length for converting a digital audio signal in units of M into frequency spectrum data on the number M of frequency axes, or a digital unit in units of number N given by M / K, where K is a positive integer in time series. A frequency conversion unit for calculating frequency spectrum data by a second conversion length for converting an audio signal into N frequency spectrum data on a frequency axis where the number is continuous with K on the time axis; A conversion length determining unit for selecting a first conversion length or a second conversion length indicating a unit for calculating frequency spectrum data in accordance with the digital audio signal;
A minimum audible value supply unit 44 for supplying a minimum audible value on the frequency axis according to the first conversion length or the second conversion length;
Is a human auditory sense used to determine the number of bits for quantization of the frequency spectrum data based on the frequency spectrum data calculated by the frequency conversion unit 41 and the conversion length selected by the conversion length determination unit 42. An auditory model determining unit 55 that calculates an auditory model corresponding to the characteristic determines the number of bits when quantizing the frequency spectrum data using the auditory model calculated by the auditory model determining unit 54, and quantizes the frequency spectrum data. It is a quantization and coding unit that converts the data into a coded bit string. The encoding process is M
This is performed in units of performing the encoding process on the digital audio signals.

Here, the reason why the frequency conversion is performed by the frequency conversion unit 41 using two different conversion lengths of the first conversion length or the second conversion length will be described below with reference to FIG. When a digital audio signal encoded by the above-described transform encoding method is decoded to reproduce the digital audio signal, a quantization error determined by the number of quantization bits at the time of encoding appears dispersedly in units for performing frequency conversion. . That is, the digital audio signal whose signal level changes greatly is
In the case of encoding with the conversion length of
As shown in (a), a quantization error that appears at the same time as a high-level signal is not perceived, but a quantization error that appears at the same time as a low-level signal is perceived. In order to prevent this, the unit for performing the frequency conversion by selecting the second conversion length is reduced to shorten the time during which the quantization error appears, as shown in FIG. It is difficult to be done. By performing the frequency conversion by selecting the conversion length in this manner, in the encoding process performed in units of M digital audio signals, when the first conversion length is selected, the M digital audio signals are converted. Frequency conversion is performed as a unit, and M pieces of frequency spectrum data Sp (0, j) (0 ≦ j ≦ M−1)
Is calculated. When the second conversion length is selected, N
Frequency conversion is performed in units of digital audio signals, and N frequency spectrum data Sp (i, j)
(0.ltoreq.i.ltoreq.K-1, 0.ltoreq.j.ltoreq.N-1) are calculated, and K frequency spectrum data groups composed of the N frequency spectrum data are continuously calculated on the time axis. FIG.
Is the frequency spectrum data Sp (i, j) (i = j) calculated when the first conversion length is selected with M = 16.
(0, 0 ≦ j ≦ 15). FIG.
Is the frequency spectrum data Sp calculated when M = 16, N = 4, K = 4 and the second conversion length is selected.
It is a figure showing a situation of (i, j) (0 ≦ i ≦ 3, 0 ≦ j ≦ 3). As shown in FIG. 7, the frequency spectrum data calculated when the second conversion length is selected is different from the frequency spectrum data calculated when the first conversion length is selected by the frequency spectrum on the frequency axis. Number of data is 1
/ K, that is, 1/4, and the frequency resolution becomes low, and the number of frequency spectrum data on the time axis becomes K times, that is, 4 times, and the time resolution becomes high.

The encoding process performed by the conventional audio signal encoding device shown in FIG. 5 will be described below. First, the conversion length determination unit 42 determines a conversion length indicating a unit for performing frequency conversion according to a time-series digital audio signal,
The determined conversion length is notified to the frequency conversion unit 41, the minimum audible value supply unit 43, the auditory model calculation unit 44, and the quantization and encoding unit 45 by the conversion length designation signal S41.

[0010] Next, in the frequency conversion unit 41, according to the conversion length designation signal S41 indicating the first conversion length or the second conversion length notified from the conversion length determination unit 42, the time-series M digital audio signals. Is frequency-converted to calculate frequency spectrum data Sp (i, j). Here, the first
When the conversion length is selected, i = 0, 0 ≦ j ≦ M−1, and when the second conversion length is selected, 0 ≦ i ≦ K−
1,0 ≦ j ≦ N−1.

In the minimum audible value supply unit 43, the minimum audible value Q1 (in the frequency axis corresponding to the first conversion length or the second conversion length) according to the conversion length designation signal S41 notified from the conversion length determination unit 42. j) or Q2 (j) is selected and supplied to the auditory model determination unit 44. Here, when the first conversion length is selected, 0 ≦ j ≦ M−1, and when the second conversion length is selected, 0 ≦ j ≦ N−1.

The auditory model calculation unit 44 converts the conversion length designation signal S41 notified from the conversion length determination unit 42 and the frequency spectrum data Sp calculated by the frequency conversion unit 41.
Based on (i, j) and the minimum audible value Q1 (j) supplied from the minimum audible value supply unit 43, it is used when determining the number of bits for quantizing the frequency spectrum data Sp (i, j). A masking threshold level M (i, j) is calculated.

In the quantization and coding unit 45, the masking threshold level M calculated by the auditory model calculating unit 44
Using (i, j), the frequency spectrum data Sp
The number of bits for quantizing (i, j) is determined, and the frequency spectrum data Sp (i, j) is quantized to output a coded bit sequence. FIG. 8 is a diagram showing the minimum audible value Q1 (j) for each frequency supplied from the minimum audible value supply unit 43 when the first conversion length when M = 16 is selected. In FIG. 8, the minimum audible value for the frequency j (0 ≦ j ≦ 15) is indicated by the shaded level Q1 (j).

For this first conversion length, FIG.
6, when N = 4 and K = 4, the minimum audible value for each frequency supplied from the minimum audible value supply unit 43 when the second conversion length at which the frequency resolution is 1 / K, that is, 1/4 is selected. Value Q
It is a figure showing 2 (j). In FIG. 9, frequency j (0
≦ j ≦ 3), the minimum audible value is the level Q of the shaded portion.
2 (j).

Conventionally, generally, Q2 (0) shown in FIG.
Is the smallest Q1 of Q1 (j) (0 ≦ j ≦ 3) in FIG.
9 is the same level as Q2 (1) in FIG. 9, and is the same level as Q1 (7) which is the minimum among Q1 (j) (4 ≦ j ≦ 7) in FIG. (2) is Q1 in FIG.
(J) (8 ≦ j ≦ 11), which is the same level as the minimum Q1 (10), and Q2 (3) in FIG.
1 (j) (12 ≦ j ≦ 15), the smallest Q1 (1
The minimum audible value was thinned out so as to be the same level as in 2).

This means that, as shown in FIG. 9, by selecting the second conversion length having a low frequency resolution, the minimum audible value Q2 () used for determining the masking threshold level M (i, j) is determined. j) is lower than the case where the first conversion length is selected. FIGS. 8 and 9 show an example in which the minimum audible value is supplied for each frequency spectrum data. However, blocks continuous on the frequency axis in which a plurality of frequency spectrum data are collected for each critical band used in psychoacoustic psychology. When encoding in units of
The minimum audible value is provided for each block. When the first transform length is selected, the number of blocks on the frequency axis is M = 1.
When the second transform length is selected, the minimum audible value is represented as shown in FIG. 8, and when the number of blocks on the frequency axis is M = 4 when the second transform length is selected, the minimum audible value is represented by FIG. It is represented as

[0017]

However, in the above-described conventional audio signal encoding apparatus, the minimum audible value when the second conversion length is selected is smaller than that when the first conversion length is selected. When the minimum audible value is used as an auditory model used to determine the number of bits to be allocated to the frequency spectrum data, the number of bits to be allocated to the frequency spectrum data when the second conversion length is selected is changed to the first conversion. The length becomes longer than when the length is selected, and the number of bits required for encoding the frequency spectrum data increases.

As described above, since the number of bits necessary for encoding the frequency spectrum data when the second transform length is selected increases, the quantization error increases due to the lack of the number of bits, and the sound quality is reduced. There was a problem that it deteriorated. The present invention solves the above-mentioned conventional problems, and can reduce the number of bits required for encoding frequency spectrum data to be executed at the time of calculating an auditory model. Provided is an audio signal encoding device and an audio signal encoding method capable of preventing deterioration of sound quality without further reducing the number of quantization bits for encoding even when the audio signal is present.

[0019]

In order to solve the above-mentioned problems, an audio signal encoding apparatus and an audio signal encoding method according to the present invention use a second transform length having a low frequency resolution when calculating an auditory model. By the selection, even when the minimum audible value becomes low, the masking threshold level indicating the auditory model is prevented from being lowered by using the time masking effect, and the number of bits allocated to the frequency spectrum data is reduced. .

As described above, it is possible to reduce the number of bits required for encoding the frequency spectrum data to be executed at the time of calculating the auditory model, and to perform encoding even when the number of bits is insufficient as a whole. Deterioration of sound quality can be prevented without further reducing the number of quantization bits.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An audio signal encoding apparatus according to a first aspect of the present invention converts a digital audio signal having a unit of number M in time series into a number M of frequency spectrum data. A digital audio signal whose unit is a conversion length or a number N given by M / K, where K is a positive integer in a time series, is converted into a number N of frequency spectrum data whose number on the time axis is continuous with K. 2, the frequency conversion unit for calculating the frequency spectrum data, the first conversion length or the second conversion length when calculating the frequency spectrum data by the frequency conversion unit, A conversion length determining unit for selecting according to the digital audio signal, and a minimum audible value for supplying a minimum audible value on a frequency axis according to the first conversion length or the second conversion length Supply unit, based on the frequency spectrum data obtained by the frequency conversion unit and the conversion length selected by the conversion length determination unit, when determining the number of bits for quantization of the frequency spectrum data An auditory model calculating unit that calculates an auditory model corresponding to the human auditory characteristics to be used, and quantizing the frequency spectrum data with the number of bits determined by the auditory model calculated by the auditory model calculating unit, to obtain an encoded bit sequence. An audio signal encoding apparatus having a quantization and encoding unit for generating the second transform length, wherein when the second transform length is selected, the number on the time axis calculated by the frequency transform unit is K
A time masking calculation unit that calculates a time masking threshold using a time masking effect between a plurality of the frequency spectrum data continuous on the time axis, the frequency spectrum data and the minimum A configuration including the auditory model determination unit that calculates an auditory model using an audible value and the temporal masking threshold.

In the audio signal encoding apparatus according to the second aspect, when the second transform length according to the first aspect is selected, L is a positive integer equal to or greater than 1 and the N frequency spectrum data are converted to the first transform length. It divides into L blocks including at least one frequency spectrum data, calculates the maximum value of the absolute value of the frequency spectrum data for each block, or the signal level obtained by the sum of squares, and calculates the number on the time axis. Has a time masking calculation unit that calculates a time masking threshold using a time masking effect between a plurality of the signal levels continuous on the time axis among the number L of the signal levels continuous with K. .

According to a third aspect of the present invention, in the audio signal encoding method, a first conversion length for converting a digital audio signal in units of the number M of time series into frequency spectrum data of the number M or a time series is used. M / K where K is a positive integer
Is selected in accordance with the digital audio signal from among the second conversion lengths for converting the digital audio signal in units of the number N given by the above into the number N of frequency spectrum data whose number on the time axis is continuous with K. Calculating the frequency spectrum data from the digital audio signal with the converted length, and determining the number of bits for quantization of the frequency spectrum data based on the frequency spectrum data and the selected conversion length. An audio for calculating an auditory model corresponding to the human auditory characteristic to be used, quantizing the frequency spectrum data calculated by the selected conversion length with the calculated number of bits determined by the auditory model, and generating an encoded bit sequence A signal encoding method, wherein the second transform length is selected in the calculation of the auditory model. Is time masking using a time masking effect between a plurality of frequency spectrum data continuous on the time axis among the frequency spectrum data whose number on the time axis calculated by the conversion length is continuous with K. A threshold value is calculated, and an auditory model is calculated using the frequency spectrum data, the minimum audible value on the frequency axis according to the first conversion length or the second conversion length, and the time masking threshold value.

According to a fourth aspect of the present invention, in the audio signal encoding method, when the second transform length according to the third aspect is selected, L is a positive integer of 1 or more, and the N number of frequency spectrum data are converted to N. It divides into L blocks including at least one frequency spectrum data, calculates the maximum value of the absolute value of the frequency spectrum data for each block, or the signal level obtained by the sum of squares, and calculates the number on the time axis. Is a method of calculating a time masking threshold using a time masking effect between a plurality of signal levels continuous on a time axis among a number L of the signal levels continuous with K.

According to these configurations and methods, when the second conversion length having a low frequency resolution is selected at the time of calculating the auditory model, even when the minimum audible value is low, the auditory sense can be obtained using the time masking effect. The masking threshold level indicating the model is prevented from lowering, and the number of bits allocated to the frequency spectrum data is reduced. Hereinafter, an audio signal encoding device and an audio signal encoding method according to embodiments of the present invention will be specifically described with reference to the drawings. (Embodiment 1) An audio signal encoding apparatus according to Embodiment 1 of the present invention will be described.

FIG. 1 is a block diagram showing the configuration of the audio signal encoding apparatus according to the first embodiment. In FIG. 1, a frequency conversion unit 1, a conversion length determination unit 2, a minimum audible value supply unit 4, and a quantization and encoding unit 5 are components that perform the same operations as the conventional audio signal encoding device in FIG. . 3 indicates that when the second conversion length is selected,
The time masking threshold T using the time masking effect between two consecutive frequency spectrum data Sp (i, j)
It is a time masking calculation unit that calculates m (i, j).
5 is the minimum audible value Q1 supplied from the minimum audible value supply unit 4.
(J) or Q2 (j) and time masking calculating section 3
In the auditory model determining unit that determines a masking threshold level M (i, j) indicating the auditory model based on the time masking threshold Tm (i, j) calculated in the above and the frequency spectrum data Sp (i, j). is there. Auditory model determination unit 5
Is a masking threshold level M indicating the auditory model based on the minimum audible value Q1 (j) supplied from the minimum audible value supply unit 4 and the frequency spectrum data Sp (i, j) when the first conversion length is selected. When (i, j) is determined and the second conversion length is selected, the minimum audible value Q2 (j) supplied from the minimum audible value supply unit 4 and the time masking threshold Tm calculated by the time masking calculation unit 3 (I, j), and the larger one is compared with the frequency spectrum data S
Based on p (i, j), a masking threshold level M (i, j) indicating the auditory model is determined and output.

FIG. 2 shows the time masking calculating section 3 shown in FIG.
FIG. 7 is a diagram showing a time masking threshold Tm (i, j) calculated using a time masking effect between K pieces of continuous frequency spectrum data Sp (i, j) on the time axis in FIG. FIG. 3 shows four continuous frequency spectrum data Sp on the time axis when {j = 0, K = 4}.
Sp (i1,0) becomes S between (i, 0) (0 ≦ i ≦ 3)
p (i2,0) (0 ≦ i1 ≦ 3,0 ≦ i2 ≦ 3, i1}
A time masking threshold Tm (i, 0) (i = 0,..., 3) based on the time masking effect given to i2) is shown.

The time masking calculation section 3 calculates a time masking threshold Tm (i, j) by a time masking effect that each frequency spectrum data Sp (i, j) gives to frequency spectrum data continuous on the time axis. , The maximum value of which is defined as a time masking threshold Tm (i, j) for the frequency spectrum Sp (i, j). Also,
In order to reduce the processing amount, a masking level between frequency spectrum data adjacent on the time axis may be calculated. Furthermore, the calculation of the masking level between the K consecutive frequency spectrum data Sp (i, j) on the time axis may be performed at an arbitrary frequency j.

The time masking threshold Tm shown in FIG.
When (0,0) is calculated by the time masking calculation unit 3, since Tm (0,3) is smaller than the minimum audible value Q2 (0) supplied from the minimum audible value supply unit 4, the auditory model determination is performed. In the section 5, a masking threshold level M used for determining the number of bits for quantizing the frequency spectrum data Sp (0,0) based on the minimum audible value Q2 (0) and the frequency spectrum data Sp (0,0). (0,0) is determined. However, Tm (1,0) is the minimum audible value Q
2 (0), Tm (1,0) and frequency spectrum data Sp
(1, 0) and the frequency spectrum data Sp
A masking threshold level M (1,0) used to determine the number of bits for quantizing (1,0) is determined. Tm
(2,0) and Tm (3,0) also have a minimum audible value Q
2 (0), masking threshold levels M (2,0) and M based on these values and frequency spectrum data Sp (2,0) and Sp (3,0).
(3, 0) is determined.

Thus, when the second conversion length is selected, the frequency spectrum data Sp continuous on the time axis is selected.
The time masking threshold Tm (i, j) due to the time masking effect during (i, j) is calculated, and when the time masking threshold Tm (i, j) is larger than the minimum audible value level Q2 (j). , This time masking threshold Tm (i,
j) and the frequency spectrum data Sp (i, j) based on a masking threshold level M (i,
j), the frequency spectrum data Sp
The number of bits allocated to (i, j) can be reduced, and the number of bits required for encoding can be reduced.

By reducing the number of bits necessary for encoding in this way, it is possible to prevent a quantization error from becoming large due to an insufficient number of bits and to prevent sound quality deterioration. In the description of the first embodiment, the time masking threshold Tm (i, i,
Although the case where the masking threshold level M (i, j) indicating the auditory model is determined based on the larger one of j) and the minimum audible value Q2 (j) and the frequency spectrum data Sp (i, j) has been described, , The minimum audible value Q2 (j) is corrected using the time masking threshold value Tm (i, j), and a masking threshold value indicating an auditory model based on the corrected minimum audible value and the frequency spectrum data Sp (i, j). Level M
(I, j) may be determined.

In the description of the first embodiment, a time masking threshold is calculated by a time masking effect between frequency spectrum data continuous on the time axis, and the minimum audible value supplied for each time masking threshold and frequency spectrum data is calculated. From the value, the frequency spectrum data Sp (i, j)
The case where the masking threshold level M (i, j) indicating the auditory model used for determining the number of quantization bits for each is described, but a plurality of frequency spectrum data Sp ( When encoding is performed in units of blocks L (L is a positive integer) obtained by combining i, j) on a frequency axis basis, the absolute value of the frequency spectrum data Sp (i, j) for each block is calculated. The signal level P (i, k) given by the maximum value or the sum of squares (0 ≦ k ≦ L
-1) to calculate a time masking threshold Tm (i, k) due to a time masking effect between signal levels P (i, k) for each block on the time axis. Minimum audible value Q3 (k) (0 ≦ k ≦ L−
The masking threshold level M (i, k) indicating the auditory model for each block used when determining the number of bits for quantizing the frequency spectrum data Sp (i, j) included in the block from 1) is determined. Even when the minimum audible value is reduced by selecting the second conversion length, the masking threshold level M (i, k) indicating the auditory model is prevented from lowering by using the time masking effect, and the frequency spectrum data is reduced. The number of bits to be allocated can be reduced. (Embodiment 2) An audio signal encoding method according to Embodiment 2 of the present invention will be described.

FIG. 3 is a flowchart showing a flow of processing of the audio signal encoding method according to the second embodiment. In FIG. 3, reference numeral 31 denotes a frequency conversion process 32, which will be described later, according to the digital audio signal.
A conversion length determination process for selecting a first conversion length or a second conversion length indicating a unit for calculating p (i, j). 32 is a process for converting a digital audio signal in units of M in time series on the frequency axis. The number M of frequency spectrum data Sp (i,
j) the first conversion length, or a digital audio signal in units of the number N given by M / K, where K is a positive integer in a time series, and a frequency axis in which the number is continuous with K on the time axis. The above number N of frequency spectrum data Sp
The frequency conversion processing for calculating the frequency spectrum data Sp (i, j) based on the second conversion length to be converted to (i, j). The frequency conversion processing 33 determines the conversion length determined in the conversion length determination processing 31 and branches. A conversion length determination process 34 is a time masking threshold by a time masking effect between frequency spectrum data Sp (i, j) continuous on the time axis when the second conversion length is selected in the conversion length determination process 31. Tm (i,
a time masking level calculation process 35 for calculating j), frequency spectrum data Sp (i, j), a minimum audible value Q1 (j) or Q2 (j) corresponding to the conversion length, and a time masking level calculation process 34 Based on the calculated time masking threshold value Tm (i, j), an auditory model used for determining the number of bits to be allocated to the frequency spectrum data Sp (i, j) in the quantization and encoding process 36 described later is described. The masking threshold level M (i,
This is an auditory model determination process for calculating j).

Here, the minimum audible value Q1 (j) or Q
2 (j) is the same as the minimum audible value supplied from the minimum audible value supply unit 43 in the conventional audio signal encoding device, and is shown in FIG. 8 and FIG. Reference numeral 36 denotes frequency spectrum data Sp (i, j) based on the frequency spectrum data Sp (i, j) and the masking threshold level M (i, j) indicating the auditory model calculated in the auditory model determination processing 35 according to the conversion length. , J) are assigned the number of quantization bits for quantizing the frequency spectrum data Sp (i, j).
j) is a quantization and encoding process that quantizes j) to generate an encoded bit stream.

In the audio signal encoding method shown in the flowchart of FIG. 3, when the second transform length is selected, the frequency mask data Sp (i, j) a masking threshold Tm (i,
j) is calculated, and in the auditory model determination processing 35, the larger one of the minimum audible value level Q2 (j) corresponding to the second conversion length and the time masking threshold Tm (i, j), and the frequency spectrum data Sp (i) , J) to determine the masking threshold level M (i, j) indicating the auditory model, it is possible to reduce the number of bits allocated to the frequency spectrum data Sp (i, j), which is necessary for encoding. Bit number can be reduced.

Thus, by reducing the number of bits required for encoding, it is possible to prevent the number of bits from becoming insufficient and the number of quantization bits from being reduced, thereby preventing sound quality deterioration. In the description of the second embodiment, the time masking threshold Tm
A case where a masking threshold level M (i, j) indicating an auditory model is determined based on the larger of (i, j) and the minimum audible value Q2 (j) and the frequency spectrum data Sp (i, j). As described above, the time masking threshold Tm (i,
j), the minimum audible value Q2 (j) is corrected, and the corrected minimum audible value and the frequency spectrum data Sp (i,
j), the masking threshold level M (i, j) indicating the auditory model may be determined.

In the description of the second embodiment, the time masking threshold is calculated by the time masking effect between the frequency spectrum data continuous on the time axis, and the minimum audible value supplied for each of the time masking threshold and the frequency spectrum data is calculated. From the value, the frequency spectrum data Sp (i, j)
The case where the masking threshold level M (i, j) indicating the auditory model used for determining the number of quantization bits for each is described, but a plurality of frequency spectrum data Sp ( When encoding is performed in units of blocks L (L is a positive integer) obtained by combining i, j) on a frequency axis basis, the absolute value of the frequency spectrum data Sp (i, j) for each block is calculated. The signal level P (i, k) given by the maximum value or the sum of squares (0 ≦ k ≦ L
-1) to calculate a time masking threshold Tm (i, k) due to a time masking effect between signal levels P (i, k) for each block on the time axis. Minimum audible value Q3 (k) (0 ≦ k ≦ L−
The masking threshold level M (i, k) indicating the auditory model for each block used when determining the number of bits for quantizing the frequency spectrum data Sp (i, j) included in the block from 1) is determined. Even when the minimum audible value is reduced by selecting the second conversion length, the masking threshold level M (i, k) indicating the auditory model is prevented from lowering by using the time masking effect, and the frequency spectrum data is reduced. The number of bits to be allocated can be reduced.

[0038]

As described above, according to the present invention, even when the second conversion length having a low frequency resolution is selected in the calculation of the auditory model, the masking level indicating the auditory model is not reduced. The number of bits allocated to frequency spectrum data can be reduced.

Therefore, it is possible to reduce the number of bits required for encoding the frequency spectrum data executed when calculating the auditory model, and to reduce the number of bits for encoding even when the number of bits is insufficient as a whole. It is possible to prevent the sound quality from deteriorating without further reducing the number of coded bits.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an audio signal encoding device according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram of frequency spectrum data according to the first embodiment.

FIG. 3 is a flowchart showing processing of an audio signal encoding method according to the second embodiment of the present invention;

FIG. 4 is a block diagram showing a configuration of a conventional audio signal encoding device.

FIG. 5 is a diagram showing a digital audio signal encoded and decoded using a first conversion length and a second conversion length in the conventional example.

FIG. 6 is an explanatory diagram of frequency spectrum data obtained with a first conversion length in the conventional example.

FIG. 7 is an explanatory diagram of frequency spectrum data obtained with a second conversion length in the conventional example.

FIG. 8 is an explanatory diagram of a minimum audible value on a frequency axis corresponding to a first conversion length in the conventional example.

FIG. 9 is an explanatory diagram of a minimum audible value on a frequency axis corresponding to a second conversion length in the conventional example.

[Explanation of symbols]

 1, 41 frequency conversion unit 2, 42 conversion length determination unit 3 time masking calculation unit 4, 43 minimum audible value supply unit 5, 44 auditory model determination unit 6, 45 quantization and coding unit

Claims (4)

[Claims] 1. A first conversion length for converting a digital audio signal in units of the number M of time series into frequency spectrum data of the number M, or given by M / K with K being a positive integer in the time series. The number N of digital audio signals whose number on the time axis is continuous with K on the time axis
A frequency conversion unit for calculating the frequency spectrum data, and a first conversion length for calculating the frequency spectrum data by the frequency conversion unit. Alternatively, a conversion length determining unit that selects a second conversion length according to the digital audio signal, and a minimum audible value that supplies a minimum audible value on a frequency axis according to the first conversion length or the second conversion length. When determining the number of bits for quantization of the frequency spectrum data based on the supply unit and the frequency spectrum data obtained by the frequency conversion unit and the conversion length selected by the conversion length determination unit. An auditory model calculation unit that calculates an auditory model corresponding to a human auditory characteristic to be used, and the auditory model is determined by the auditory model calculated by the auditory model calculator. An audio signal encoding device having a quantization and encoding unit that quantizes the frequency spectrum data by the number of bits to generate an encoded bit sequence, wherein when the second conversion length is selected, A time masking threshold is calculated by using a time masking effect between a plurality of the frequency spectrum data continuous on the time axis among the frequency spectrum data whose number on the time axis calculated by the conversion unit is continuous with K. An audio signal encoding apparatus, comprising: a time masking calculation unit that performs the calculation; and the auditory model determination unit that calculates an auditory model using the frequency spectrum data, the minimum audible value, and the time masking threshold. 2. When the second conversion length is selected, L
Divides the N pieces of frequency spectrum data into L blocks including at least one piece of frequency spectrum data as one or more positive integers, and calculates the maximum value or the square of the absolute value of the frequency spectrum data for each block. The signal level calculated by the sum is calculated, and the number on the time axis is K
And a time masking calculation unit that calculates a time masking threshold using a time masking effect between a plurality of the signal levels that are continuous on the time axis among the number L of the signal levels that are continuous in the above manner. The audio signal encoding device according to claim 1. 3. A first conversion length for converting a digital audio signal in units of the number M of time series into frequency spectrum data of the number M, or given by M / K with K being a positive integer in the time series. The number N of digital audio signals whose number on the time axis is continuous with K on the time axis
Calculating the frequency spectrum data from the digital audio signal with a conversion length selected according to the digital audio signal among the second conversion lengths to be converted into the frequency spectrum data of Based on the length, to calculate an auditory model corresponding to the human auditory characteristics used in determining the number of bits for quantization of the frequency spectrum data,
An audio signal encoding method for quantizing frequency spectrum data calculated by the selected conversion length with the calculated number of bits determined by the auditory model, and generating an encoded bit sequence, wherein the audio model includes When the second conversion length is selected, among the frequency spectrum data whose number on the time axis calculated by the conversion length is continuous with K, among the plurality of frequency spectrum data continuous on the time axis, A time masking threshold using the time masking effect of the above, and using the frequency spectrum data, a minimum audible value of the frequency axis corresponding to the first conversion length or the second conversion length, and an auditory sense using the time masking threshold An audio signal encoding method comprising calculating a model. 4. When the second conversion length is selected, L
Divides the N pieces of frequency spectrum data into L blocks including at least one piece of frequency spectrum data as one or more positive integers, and calculates the maximum value or the square of the absolute value of the frequency spectrum data for each block. The signal level calculated by the sum is calculated, and the number on the time axis is K
4. The audio according to claim 3, wherein a time masking threshold is calculated using a time masking effect between a plurality of signal levels that are continuous on a time axis among the number L of the signal levels that are continuous. Signal encoding method.