JPH0918348A - Acoustic signal encoding device and acoustic signal decoding device - Google Patents

Abstract

PURPOSE: To improve a 'level' that the increase of sub information exceeds the compressibility of main information and to further improve the compressibility of the entire output signal string. CONSTITUTION: This acoustic signal encoding device for converting acoustic signals 1 to the digital code information of high quality and outputting it as encoded data is provided with a band divider 2 for dividing the acoustic signal 1 into plural frequency components, an adaptive sub information encoder 41 for encoding a scale factor obtained from the output of the band divider 2 and noise allocation information obtained from an auditory model 4, generating the sub information and generating a sub information code 45 from a sub information prediction signal, an abstract audition model 42 for generating the sub information prediction signal from the encoded sub information, a main information encoder 8 for encoding the frequency component 3 outputted from the band divider 2 based on the sub information 45 encoded in the adaptive sub information encoder 41 and generating a main information code 9 and a code synthesizer 10 for synthesizing the main information code 9 and the sub information code 45 and generating digital code information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is for converting wide band analog audio information into high quality digital code information, or for converting high quality digital code information back into wide band analog audio information. The present invention relates to an improved technique of an acoustic signal encoding device or an acoustic signal decoding device used.

[0002] In recent years, there has been an increasing interest in "multimedia technology" capable of centrally handling not only textual information but also almost all human-recognizable information such as image information and audio information. However, as the amount of information has increased significantly, the transmission cost (typically the communication cost using a telephone line) and the efficiency of storage in storage media have hindered the spread of the information. A compression technique and a decompression technique that can reduce the amount of information with contents close to the original information are required. The present invention relates to such multimedia technology, in particular to compression and decompression technology for wideband analog audio information. The term “audio” is a word generally used in the field of music, but in the present specification, all information that can be recognized by hearing (however, whether or not the information is in the audible range) is used. Shall be pointed out.

[0003]

2. Description of the Related Art As compression and decompression techniques for wideband analog audio information (hereinafter sometimes referred to as "acoustic signal"), the following techniques have been known. "First conventional example" A method of compressing (reducing) the amount of information after encoding by predicting a waveform following a known portion of an acoustic signal waveform. For example, DPCM and ADPCM are typical examples. In these representative examples, the difference between the peak value of the point of interest of the acoustic signal waveform and the peak value immediately before it is calculated, and the difference information is transmitted or stored to reduce the total amount of information. "Second conventional example" Hearing frequency masking effect ...
............................, using using the effect that the frequency component B of a lower level than A, which is located near the frequency component A, is masked by the human ear and cannot be heard. The process of the first conventional example is separated into a plurality of frequency bands and performed. For example, ITU
G. (International Telecommunication Union) In the 722 standard, the acoustic signal is 4
It is divided into two frequency bands, and A for each band component
You are using DPCM. "Third conventional example" A precise model of the frequency masking effect is applied in order to use the auditory properties to the limit. For example, according to the ISO / IEC 11172.3 international standard (MPEG Audio), the number of divisions of the frequency band is increased from 32 to 384, and the optimum code amount is assigned to each band. ACTS (American Communications Commission) standard audio transmission (Dolby AC3) and MiniDi
A similar method is adopted for the sc method and the like.

Of these three conventional examples, the third conventional example has the highest compression efficiency. FIG. 7 is a block diagram of an encoder according to the third conventional example. In this figure, 1
Is an input acoustic signal, 2 is a band divider that divides the input acoustic signal 1 into a plurality of frequency components 3 without losing information,
4 is a noise allocation by analyzing the input acoustic signal 1 (amount representing the maximum amplitude of noise that does not cause deterioration in hearing characteristics) 5
, 6 is a sub-information encoder that generates sub-information code 7 based on the plurality of frequency components 3 and noise allocation 5, and 8 is a main information based on the plurality of frequency components 3 and sub-information code 7. Main information encoder for generating information code 9, 10
Is a code string generator that synthesizes the sub information code 7 and the main information code 9 to generate an output code 11 for transmission or storage.

FIG. 8 is a block diagram of a decoder according to the third conventional example. In this figure, 21 is an input code (see FIG.
Of the input code 21), 22 is a code decomposer that decomposes the input code 21 into main information code 23 and sub information code 24, and 25
Is from the sub information code 24 to the sub information 26 such as magnification and quantization accuracy.
, 27 is a main information decoder that decodes the main information code 23 into a plurality of frequency components 28 based on the sub information 26, and 29 is a combination of the plurality of frequency components 28 and an original acoustic signal 30. Is a band synthesizer for reproducing the.

The encoder shown in FIG. 7 variably obtains the scaling factor and quantization precision of the input acoustic signal divided for each frequency and time block, and the waveform itself is determined by these scaling factor and quantization precision. It is compressed to the minimum digital code amount and output. The output code string includes “main information” and “sub information”, and the main information is the main component of the information, but the sub information is
It is auxiliary information necessary for decoding the main information, such as magnification and quantization accuracy. By giving the sub-information more finely, the coding efficiency (compression rate) of the main information can be increased.

[0007]

However, in the third conventional example, if the fineness of the sub information is increased in order to increase the compression rate of the main information, the "sub" in the output code string is increased. The information itself has the contradictory effect of increasing the amount of information, and the sub-information can be increased in fineness only to the "level" at which the increase in the sub-information does not exceed the compression rate of the main information. There is a problem that the compression rate of the entire column cannot be increased.

Therefore, an object of the present invention is to increase the "level" at which the increase amount of the sub information exceeds the compression ratio of the main information, thereby further improving the compression ratio of the entire output signal sequence.

[0009]

According to the first aspect of the present invention,
An acoustic signal encoding device for converting an acoustic signal into high-quality digital code information and outputting it as encoded data,
A band divider for dividing the acoustic signal into a plurality of frequency components, a scale factor obtained from the output of the band divider, and noise allocation information obtained from an auditory model are encoded to generate sub information, and sub information An adaptive sub-information encoder that generates a sub-information code from a prediction signal, an abstract auditory model that generates a sub-information prediction signal from encoded sub-information, and sub-information encoded by the adaptive sub-information encoder A main information encoder that encodes a frequency component output from the band divider based on the above to generate a main information code, and a code synthesizer that synthesizes the main information code and the sub information code to generate digital code information And are provided.

According to a second aspect of the present invention, in the first aspect of the present invention, noise allocation information is generated so that the input acoustic signal is analyzed and coding can be performed without causing deterioration in human auditory characteristics. It is characterized by having an auditory model that can be used. According to a third aspect of the present invention, in the first aspect of the invention, the adaptive side information encoder uses a statistical property of side information that is predicted using a side information prediction model such as an abstract auditory model. It is characterized by adaptively encoding the sub-information.

According to a fourth aspect of the present invention, there is provided an acoustic signal decoding device which inputs digital code information as encoded data, inversely converts the encoded data into an acoustic signal and outputs the acoustic signal.
An abstract auditory model that generates a sub information prediction signal from sub information in digital code information, an adaptive sub information decoder that decodes the sub information code using the sub information prediction signal, and based on the decoded sub information , A main information decoder for decoding the main information code in the digital code information to return it to a frequency component, and a band synthesizer for returning the frequency component to the original acoustic signal.

According to a fifth aspect of the invention, in the invention according to the fourth aspect, the adaptive side information decoder determines the statistical property of the side information predicted by using a side information prediction model such as an abstract auditory model. It is characterized in that it utilizes and adaptively decodes the sub information that is adaptively encoded.

[0013]

In the present invention, a sub information prediction signal having a smaller amount of information than the sub information is created using the abstract auditory model, and the sub information code and the main information generated from this sub information prediction signal are combined. Digital code information is generated. Therefore,
The amount of sub information in the digital code information decreases,
As a result of the increase in the "level" at which the increase amount of the sub information exceeds the compression ratio of the main information, the compression ratio of the entire output signal sequence can be further improved.

[0014]

Embodiments of the present invention will be described below with reference to the drawings. 1 to 6 are diagrams showing an embodiment of an acoustic signal encoding device and an acoustic signal decoding device according to the present invention. In addition, in each drawing, the same components as those of the conventional example are denoted by the same reference numerals. [Encoder] First, the configuration of the encoder of the present embodiment will be described. In FIG. 1, an input acoustic signal 1 is obtained by uniformly PCM-encoding wideband analog audio information at a constant sampling rate, and for example, a 16-bit (bit) linear PCM signal can be used. Needless to say, the optimum bit number is not limited to this bit number and may be selected according to the required accuracy. It should be noted that in the following, a one-channel (that is, monaural) acoustic signal is taken as an example, but this is for simplification of description, and a multi-channel acoustic signal is not excluded.
The input acoustic signal 1 is the following signal.

X [t]: t = 0..∞ The input acoustic signal 1 is input to the band divider 2 and the auditory model 4. The band divider 2 inputs the input acoustic signal 1 into a plurality of frequency components 3 without losing the information of the input acoustic signal 1.
It has a function of dividing into. Such a function is, for example, DFT (Discrete Fourior Transform), DHT.
(Discrete Hartley Transform), DCT (Discrete C
osine Transform) or MDCT (Modified Discrete C)
It can be realized by various known methods such as osine transform). Although not particularly limited, MDCT is used in this embodiment. In MDCT, frequency components are obtained as follows.

Input: x [m] [i] m = 0,1, ..., ∞ (segment number) i = 0, .., L-1 (sample number in segment) Output: c [m] [ k] = Σ (x [m-1] [i] × M [i, k] + x [m] [i] × M [i + L, k]) for i = 0, ..., L- 1 MDCT coefficient: M [i, k] = cos {(2k + 1) × (2i + n0) × Pi / 2 / L} n0 = (L + 1) / 2 For example, for a sine wave with a certain frequency If so, this conversion causes the output to concentrate on the components shown in Table 1 below. According to Table 1, the effect of dividing the band into L equal parts is obtained by the conversion.

[0017] The frequency component 3 is the following signal.

C [m] [k]: k = 0..L, m = 0..∞ The frequency component 3 is divided into segments for later efficient coding with side information. . The segmentation may be performed in both the time direction and the frequency direction. In the following description, the segment (b, p) is assumed to include the range m = m (b) .. m (b + 1) -1 k = k (p) .. k (p + 1) -1. .

The auditory model 4 analyzes the input acoustic signal 1 and generates noise allocation information 5 representing the maximum amplitude of noise that does not cause deterioration in auditory characteristics. For noise allocation, for example, ISO / IEC 11172.
3 Efficient coding principles defined by international standards, etc ...
········································································································· and ····························· From assigning permissible quantization noise for each band optimally according to an input signal. Here, the noise allocation is a quantity that represents the maximum amplitude allowed for the quantization noise in the time / frequency segment, for example, n [b] [p]: b = 0..∞, p = 0..M Signal. n [b] [p] is the maximum amplitude of noise in each block that does not cause deterioration in auditory characteristics. As a concrete method of noise allocation, for example, P described in ISO / IEC 11172.3 is used.
sycho Acoustic Model 2 can be applied. The allowable noise can be obtained as a function of frequency and time, and can be easily converted into the above n [b] [p].

The frequency component 3 is input to the main information encoder 8 and also to the sub information processing section 40. The sub-information processing unit 40 includes an adaptive sub-information encoder 41 and an abstract auditory model 42. The adaptive sub-information encoder 41 generates the sub-information 7 from the noise allocation information 5 and the frequency component 3 and then outputs the sub-information. The information 7 is adaptively and efficiently coded by the function of the abstract auditory model 42 to generate the sub information code 45. The adaptive sub-information encoder 41 can obtain the scale factor s [b] [p], which is the maximum amplitude of one segment, from the input frequency component c [m] [k], for example.

S [b] [p] = max (c [m] [k], {m = m (b) .. m
(B + 1) -1, k = k (p) .. k (p + 1) -1}) In addition, the required number of quantization steps Q [b] [p] is calculated from the noise allocation and the above scale factor. You can ask. Q [b] [p] = s [b] [p] / n [b] [p] By using these side information, the main information can be efficiently coded as described later, but the side information itself Since it is also necessary to encode the sub information, it is necessary to apply an efficient encoding technique also to this sub information in order to reduce the overall code amount.

Considering Q [b] [p] and s [b] [p], the following can be understood. (1) There are many segments with Q [b] [p] <1. For these segments, neither main information nor sub-information needs to be encoded (however, information indicating that Q [b] [p] <1 is required). (2) Q [b] [p] that are close to each other are not completely independent, and have the following constraints on each other. (2-1) Time Masking Effect At a certain frequency, a relatively small amplitude component existing immediately before or after a large amplitude component is masked and cannot be heard by the human ear. In Figure 3, Q [b-1] [p]
Has a substantially mountain-shaped time masking effect as shown by the dotted line. Therefore, for Q [b] [p] adjacent to Q [b-1] [p], only the portion exceeding the masking effect needs to be encoded. (2-2) Frequency masking effect On the frequency axis, a relatively small amplitude component existing in the vicinity of a certain frequency component having a large amplitude is masked and cannot be heard. In FIG. 4, Q [b] [p-1] has a substantially mountain-shaped frequency masking effect as shown by the dotted line. Therefore,
Regarding Q [b] [p] adjacent to Q [b] [p-1], only the portion exceeding the masking effect may be encoded.

By using these properties (1) and (2), the expected distribution of Q [b] [p] can be obtained when encoding a certain value of Q [b] [p], It is possible to efficiently encode Q [b] [p]. As described above, the adaptive sub-information encoder 41 adaptively encodes the sub-information by the correlation between the sub-information, and the sub-information for that purpose can be predicted by the abstract auditory model 42.

The adaptive sub-information encoder 41 inputs the roughly encoded sub-information 43 to the abstract auditory model 42 and obtains the sub-information prediction 44 from the abstract auditory model 42. As will be described later, the abstract auditory model 42 is also included in the decoder of the present invention, but since the input is the roughly encoded side information 43, the abstract auditory model on the encoding side and the abstract auditory model on the decoding side are common. Can be converted.

The abstract auditory model 42 matches the auditory model 4 in that it is a model that equivalently calculates auditory characteristics, but is different in the following points. (A) The auditory model 4 takes all of the acoustic signal 1 as an input variable, but the abstract auditory model 42 takes only the roughly encoded sub-information 43 as an input variable. (B) Although the auditory model 4 is generally treated as having an emphasis on accuracy and an increase in the amount of calculation is unavoidable, since the abstract auditory model 42 is also required at the time of decoding, the amount of calculation is emphasized (small amount of calculation) rather than accuracy. (C) The accuracy of the auditory model 4 directly affects the performance, but the abstract auditory model 42 need not be faithful to the auditory characteristics.

That is, since the abstract auditory model 42 is not intended for accurate noise allocation, it does not have to be as precise as the auditory model 4. For example, the sub information can be predicted by the following simplified method. can do.
Note that FIG. 5 shows a conceptual flow of the method. p_zero [b] [p] = 0.5 s_max [b] [p] = n1 [b] [p-1] + off1 s_min [b] [p] = max (s1 [b-1] [p] + off21, s1 [b] [p-1] + off
22) + off23 n_max [b] [p] = n1 [b] [p-1] + off3 n_min [b] [p] = max (s1 [b-1] [p] + off41, s1 [b] [ p-1] + off
42) Here, n1 [b] [p] = s1 [b] [p] / Q1 [b] [p] Predicted values have the following meanings.

P_zero [b] [p]: Probability that main information allocation is zero s_max [b] [p]: Maximum value of scale factor s_min [b] [p]: Minimum value of scale factor n_max [b] [ p]: Maximum noise value n_min [b] [p]: Minimum noise value n1 [b] [p] Noise allocation for sub information code s1 [b] [p] Scale factor for sub information code (maximum amplitude ) Off1, off21, off22, off3, off41, off42 are constants. This simple prediction method is valid for the following reasons.

S_max [b] [p] = n1 [b] [p-1] + off1 Due to the maximum amplitude of the segment (b, p), the adjacent segment (b, p-1) should be masked. Therefore, the maximum amplitude that the segment (b, p) can take can be estimated by the noise allocation of the segment (b, p-1). s_min [b] [p] = max (s1 [b-1] [p] + off21, s1 [b] [p-1] + off
22) Segments (b-1, p) and (b, p-1) both mask segment (b, p), so the lower bound of noise allocation can be estimated from them. Therefore, segment (b, p)
Must have a maximum amplitude greater than at least that level in order to be zero-allocated.

N_max [b] [p] = n1 [b] [p-1] + off3 The noise allocation of the segment (b, p) has continuity with the adjacent segment (b, p-1). , It is possible to estimate the maximum value. n_min [b] [p] = max (s1 [b-1] [p] + off41, s1 [b] [p-1] + off
42) The minimum value of noise allocation of segment (b, p) can be estimated by masking with adjacent segments.

Although these estimations include some errors, since the margins are provided by the respective offsets, the actual value of the sub-information is always contained within these maximum and minimum values. By using these maximum and minimum values, the sub information code can be generated as follows. n1 [b] [p] = Quant2 (n [b] [p] / n_min [b] [p]) s1 [b] [p] = Quant2 (s [b] [p] / s_min [b] [p ]) Here, Quant2 is a logarithmic compression type quantization function suitable for quantizing scale and noise allocation as follows.

Quant2 (x) = rint (log (x) / log (2) × a1) a1 is a constant (for example, 3.0 can be used) The side information code is immediately dequantized, and it is referred to as side information. To do. n2 [b] [p] = Requant2 (n1 [b] [p]) × n_min [b] [p] s2 [b] [p] = Requant2 (s1 [b] [p]) × s_min [b] [ p] Requant2 (x) is a function that dequantizes the integer quantized by Quant2 (x).

Requant2 (x) = Exp (x × a1) a1 is a constant and has the same value as Quant2 (x) (for example, 3.
0) is used. The above is an example of a specific method of the abstract auditory model. To perform encoding based on the information estimated by such an abstract auditory model, for example, “arithmetic encoding”, “Huffman encoding”, "Vector coding"
Alternatively, various methods such as “table reference coding” can be applied.

The main information encoder 8 receives the sub information 7 and the frequency component 3, and outputs the main information code 9. From the sub-information 7, it is possible to obtain the required S / N ratio (signal-to-noise ratio) and maximum amplitude for each frequency band. Generally, if a frequency band and a necessary S / N ratio are given, the signal can be encoded at a predetermined bit rate by various methods.

For example, the main information encoder 8 can perform encoding by the following method. First, the scale s [b]
From [p] and noise allocation n [b] [p], the number of quantization steps q [b] [p] is calculated as follows. q [b] [p] = s [b] [p] / n [b] [p] Next, the number of quantization steps q [b] [p] and scale s [b] [p]
Quantization and dequantization are defined as follows using.

Quant (x) = rint (x / s [b] [p] × q [b] [p]) Requant (x) = x / q [b] [p] × s [b] [p] rint (x) is a rounding function whose value is the integer closest to x. The main information encoder 8 performs the following quantization using this quantization function, and the code string generator 10 receives the main information code 9 and the sub information code 15 as input and finally outputs the output code 11 Is output.

Y [m] [k] = Quant (c [m] [k]) "Decoder" Next, the configuration of the decoder of this embodiment will be described.
In FIG. 2, the input code 21 is the output (output code 11) of the encoder. The code decomposer 22 uses the input code 21
Is decomposed into main information code 23 and sub information code 24. The main information code 23 is, for example, an integer quantized and scaled as follows.

Q [m] [k] m = 0..∞, k = 0..L The sub information code 24 is, for example, an integer string for reproducing the sub information for each segment as follows. s2 [b] [p] b = 0..∞, p = 0..M n2 [b] [p] b = 0..∞, p = 0..M The decoding processing unit 50 performs adaptive side information decoding. The adaptive information decoder 51, which includes the device 51 and the abstract auditory model 52, uses the same technique as the encoder to generate the sub information 36 from the sub information code 24.

N2 [b] [p] = Requant2 (n1 [b] [p]) × n_min [b] [p] s2 [b] [p] = Requant2 (s1 [b] [p]) × s_min [ b] [p] Therefore, the sub information 36 that is exactly the same as that sent to the main information encoder 8 of the encoder is also reproduced by the decoder and sent to the main information decoder 27. The sub-information 36 has the following signals.

N2 [b] [p] b = 0..∞ p = 0..M s2 [b] [p] b = 0..∞ p = 0..M Abstract auditory model 52, approximate coding sub The information 53 and the side information prediction 54 are exactly the same as the corresponding portions of the encoder, respectively, and when processing the same signal, the respective signals are also identical. Therefore, the description relating to them shall refer to the same part of the encoder.

The main information decoder 27 inputs the sub information 36 and the main information code 23, and outputs the frequency component 28. This operation forms a pair with the main information encoder 8 of the encoder, and various methods can be adopted like the main information encoder 8 and any method may be used. For example, if the main information encoder 8 of the encoder uses uniform quantization as described above, the main information decoder 27 of the decoder can obtain the frequency component 28 by inverse quantization as follows. it can.

C [m] [k] = Requant (y [m] [k]) m = 0..∞, k = 0..L-1 The band synthesizer 29 receives the frequency component 28 as an input, The output acoustic signal 30 in the form of a time series sample is converted and output. This conversion is the reverse of the operation of the band divider 2 in the encoder. For example, MDCT (Modified Discr
If you use ete cosine transform), band synthesizer 2
9 shows the reverse of IMDCT (Inverse Modified Discre
te Cosine Transform). The conversion formula of IMDCT is shown below.

Input: c [m] [k] m = 0,1, .., ∞ (segment number) k = 0,1, .., L (coefficient number in segment) y [m] [n] = 2 / N × h (n) × Σ (c [m] [k] × cos ((2k + 1) (n + n0) Pi / 2 / N)) for n = 0,1,2 .., 2N -1 n0 = N / 2 + 1/2 Output: x [m] [n] = y [m-1] [n + N] + y [m] [n] m = 0,1, .., ∞ (Segment number) n = 0,1, .., N-1 (Sample number in segment) Output is time series data.

As described above, the sub information coding method in the present invention is not affected by the main information quantization method, band division method, or the like. Therefore, the present invention can be applied regardless of the difference between various encoding methods and frequency division methods based on auditory psychology. Although a simple abstract auditory model is taken as an example for convenience of explanation, it is clear that the closer the side information predicted by the abstract auditory model to the actual side information, the higher the encoding efficiency. Therefore, the abstract auditory model may have a structure close to that of the auditory model used in the encoder. For example, the abstract auditory model and the auditory model may be the same. "Another example of the abstract auditory model" Although the example of the abstract auditory model has already been described, in addition to this, the representative value of the side information is first accurately encoded and the side information between them is predicted. Therefore, a method of encoding with a small code amount can also be used.

For example, as shown in FIG. 6, first, the scale factor is encoded every eight segments with very fine precision (8 bits). The intermediate range (every 4 segments) of the scale factor can be estimated to some extent by the abstract auditory model from the previously encoded scale factor, so the code amount (6 bits) is smaller than the previous scale factor. Can be encoded. Further, since the middle of them (every two segments) can be accurately predicted by narrowing the frequency interval from the already coded scale factor, it is possible to code with a smaller code amount (4 bits). Further, the code amount of the encoded scale factor can be reduced by more accurately predicting the intermediate scale factor from the front and the back. The same applies to noise allocation. Note that the number of segments and the number of bits are examples.

[0045]

According to the present invention, the amount of sub information in the digital code information can be reduced. Therefore, in order to increase the compression rate of the main information, even if the fineness of the sub information is increased, the amount of information of the "sub information" itself in the output code string does not increase compared to the conventional example. It is possible to increase the "level" in which the increase amount of P exceeds the compression ratio of the main information, and it is possible to further improve the compression ratio of the entire output signal sequence.

[Brief description of the drawings]

FIG. 1 is a block diagram of an encoder according to an embodiment.

FIG. 2 is a block diagram of a decoder according to an embodiment.

FIG. 3 is a graph of time masking effect.

FIG. 4 is a graph of frequency masking effect.

FIG. 5 is a main algorithm according to an embodiment.

FIG. 6 is another main part algorithm of the embodiment.

FIG. 7 is a block diagram of a conventional encoder.

FIG. 8 is a block diagram of a conventional decoder.

[Explanation of symbols]

 1: Input acoustic signal 2: Band divider 3: Frequency component 4: Auditory model 5: Noise allocation 6: Sub information encoder 7: Sub information 8: Main information encoder 9: Main information code 10: Code string generation Device 11: Output code 21: Input code 22: Code decomposer 23: Frequency component 27: Main information decoder 28: Main information code 29: Band synthesizer 30: Output acoustic signal 36: Sub information 41: Adaptive sub information encoding 42: Abstract auditory model 43: Approximately encoded side information 44: Side information prediction 45: Side information code 51: Adaptive side information decoder 52: Abstract hearing model 53: Approximately encoded side information 54: Side information prediction

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Masaya Konishi 3-12 Moriya-cho, Kanagawa-ku, Yokohama, Kanagawa Japan Victor Company of Japan, Ltd.

Claims (5)

[Claims] 1. An audio signal encoding device for converting an audio signal into high-quality digital code information and outputting it as encoded data, comprising: a band divider for dividing the audio signal into a plurality of frequency components; An adaptive sub-information encoder that generates side information by encoding the scale factor obtained from the output of the band divider and the noise allocation information obtained from the auditory model and side information, and the sub information code from the side information prediction signal, An abstract auditory model for generating a sub information prediction signal from encoded sub information, and a frequency component output from the band divider based on the sub information encoded by the adaptive sub information encoder An audio signal encoding device comprising: a main information encoder that generates an information code; and a code synthesizer that synthesizes a main information code and a sub information code to generate digital code information. apparatus. 2. An auditory model capable of generating noise allocation information so that an input acoustic signal can be analyzed and encoding can be performed without causing deterioration in human auditory characteristics. Item 1. The acoustic signal encoding device according to item 1. 3. An adaptive sub-information encoder that adaptively encodes sub-information using the statistical property of sub-information predicted using a sub-information prediction model such as an abstract auditory model. The acoustic signal encoding device according to claim 1. 4. An acoustic signal decoding device for inputting digital code information as encoded data, inversely converting the encoded data into an acoustic signal and outputting the acoustic signal, wherein the sub information predictive signal is derived from sub information in the digital code information. , An adaptive sub-information decoder that decodes the sub-information code using the sub-information prediction signal, and decodes the main information code in the digital code information based on the decoded sub-information. An audio signal decoding apparatus, comprising: a main information decoder for returning the frequency component to a frequency component; and a band synthesizer for returning the frequency component to the original acoustic signal. 5. An adaptive sub-information decoder utilizes adaptive statistical information of the sub-information, which is predicted using a sub-information prediction model such as an abstract auditory model, to adaptively code the sub-information. The acoustic signal decoding device according to claim 4, wherein the acoustic signal decoding device performs adaptive decoding.