KR101325760B1 - Apparatus and method for audio codec - Google Patents

Abstract

The present invention relates to audio signal processing, which assigns the converted frequency coefficient to each track according to a track structure determination unit for determining a track structure using a frequency coefficient converted from an audio signal and a track structure determined by the track structure determination unit. A track structure and a frequency coefficient in a conventional sinusoidal quantization scheme using a fixed track structure by an encoding device, comprising a frequency coefficient allocator and a quantizer for quantizing at least one pulse in the track based on the assigned frequency coefficients. If the pulse distributions of the two do not match, all the pulses cannot be quantized, thereby preventing sound degradation.

Description

Apparatus and method for audio codec}

TECHNICAL FIELD The present invention relates to audio signal processing, and more particularly, to a decoding and encoding technique of an audio / audio signal processing apparatus.

The present invention is derived from a study conducted as part of the IT growth engine technology development project of the Ministry of Knowledge Economy. [National Management No.: 2008-S-011-02, Title: FMC acoustic fusion codec and control technology research (standardization) Link)]

The analog speech signal is sampled and linear quantized to obtain a PCM signal. However, since the capacity of the PCM signal is large, it is difficult to store, transmit, and reproduce the data without being compressed. Accordingly, development of a voice / audio codec for compressing and reconstructing PCM signals has been in progress. Recently, voice / audio codecs use a method of quantizing an input signal from a time domain to a frequency domain.

Typical quantization methods include Tree-structured Quantization, Product Quantization, Lattice Quantization, Predictive Quantization, Address Quantization, Fine-coarse Quantization, Multistage Quantization, Trellis-coded Quantization, and Pyramid Quantization.

Product quantization is a method of quantizing each subband by dividing the transformed frequency coefficients into a plurality of subbands. In general, Product quantization takes a scalar quantization after obtaining a gain of subband frequency coefficients and vector quantizes the shape of the frequency coefficients. However, when the distribution of the frequency coefficients in the sub band is in the form of a pulse that increases in a specific position, there is a limit in expressing an accurate pulse shape by vector quantization.

A solution to this problem is sinusoidal quantization, which divides frequency coefficients into multiple tracks, or subbands, and selects pulses in the order of the largest absolute value of the coefficients in each track to determine the position and magnitude of the pulse. It is a method of quantization.

The present invention is derived from this background, and aims to further increase the sophistication of sinusoidal quantization.

The technical problem is a track structure determination unit for determining a track structure using a frequency coefficient converted from an audio signal and a frequency coefficient allocation unit for allocating the converted frequency coefficient to each track according to the track structure determined by the track structure determination unit; And a quantization unit for quantizing at least one pulse in a track based on the assigned frequency coefficients.

In addition, the technical problem is an inverse quantizer for outputting the recovered pulse parameters by inverse quantization of the quantized pulse parameters included in the input bit stream, track structure determination unit for determining the track structure based on the track parameters included in the input bit stream, It is also achieved by a decoding apparatus comprising a pulse generator for generating a pulse based on the determined pulse parameter and a coefficient generator for generating a frequency coefficient based on the determined track structure and the generated pulse.

On the other hand, calculating an energy concentration from a frequency coefficient converted from an audio signal, determining a track structure based on the calculated energy concentration, assigning the frequency coefficients to the track based on the determined track structure, A method of encoding an audio signal, comprising: selecting at least one pulse in a track and quantizing the selected pulse in the assigned track.

And determining a track structure based on the track parameters included in the input bit stream, restoring a pulse parameter by inverse quantization of the input bit stream, generating a pulse from the recovered pulse parameter, and determining and generating the track structure. It is also achieved by a method for decoding an audio signal, comprising generating and outputting a frequency coefficient on the basis of the pulse.

According to the present invention, in the existing sinusoidal quantization method using a fixed track structure, if the pulse structure of the track structure and the frequency coefficients do not coincide with each other, it is possible to prevent sound quality degradation from occurring because all pulses are not quantized.

That is, a sequential track structure is used for a signal in which sine pulses are evenly distributed over the entire band in the distribution of frequency coefficients. It can express, and it has the effect of making sound quality better.

1 is a block diagram of a general audio signal processing apparatus;
2 is a block diagram of an audio signal processing apparatus according to an embodiment;
3 is a flowchart of a method of encoding an audio signal, according to an exemplary embodiment.
4 is a flowchart of a method of decoding an audio signal, according to an exemplary embodiment.

The foregoing and further aspects of the present invention will become more apparent through the preferred embodiments described with reference to the accompanying drawings. Hereinafter, the present invention will be described in detail so that those skilled in the art can easily understand and reproduce the present invention through these embodiments.

1 is a block diagram of a general audio signal processing apparatus.

As shown in FIG. 1, a general audio signal device includes an encoder 100 and a decoder 110.

The encoder 100 generates an bitstream by encoding a quantization index of frequency coefficients. The bit string is transferred to another terminal device through a storage medium or a communication channel. The encoder 100 includes a frequency converter 102 and a quantizer 104.

The frequency converter 102 converts a voice / audio signal, that is, an input signal, from the time domain to the frequency domain. The quantizer 104 obtains a bit string by quantizing the frequency coefficients converted by the frequency converter 102, and outputs a quantization index of the frequency coefficients in the form of a bit stream.

In addition, the quantization unit 104 determines the quantization standard for each frequency according to the masking information, and quantizes the frequency coefficient according to the determined quantization standard. The quantization unit 104 may use various quantization techniques such as prediction quantization in order to improve quantization performance.

Meanwhile, the decoder 110 extracts a quantization index from the received bit string, obtains a frequency coefficient, and then converts the frequency coefficient into a time domain to restore the final audio signal.

In detail, the decoder 110 includes an inverse quantizer 112 and an inverse transform unit 114. The inverse quantization unit 112 obtains a frequency coefficient from an input bit string. The inverse transformer 114 converts a frequency coefficient obtained by the inverse quantizer 112 into a time domain and outputs an audio signal.

2 is a block diagram of an audio signal processing apparatus according to an exemplary embodiment.

As shown, an audio signal processing apparatus according to an exemplary embodiment includes a calculator 200, a track structure determiner 210, a frequency coefficient assigner 220, a pulse determiner 230, a quantizer 240, and multiplexing. The unit 250 includes a demultiplexer 270, a dequantizer 282, a pulse generator 288, and a coefficient generator 290.

The calculator 200 receives the frequency coefficients converted from the voice / audio signal and calculates energy concentration in each track structure. For example, the track structure includes a sequential structure and an interleave structure. If the number of tracks is 4, the number of pulses in each track is 2, and the frequency coefficient is 64, the energy concentration EC of each track structure can be expressed as follows.

는 절대값이 가장 큰 2개의 주파수 계수를 의미한다. 만약, 각 트랙에서 펄스의 수가 4개인 경우, 수학식 1 중

부분은

으로 변경된다.

는 절대값이 가장 큰 4개의 주파수 계수를 의미한다. 이와 같이, 산출부(200)는 펄스의 수에 기초하여 에너지 집중도를 계산할 수 있다.
here,

Denotes the two frequency coefficients with the largest absolute values. If the number of pulses in each track is four, Equation 1

Part

Is changed to

Denotes the four frequency coefficients with the largest absolute values. As such, the calculator 200 may calculate the energy concentration based on the number of pulses.

The calculator 200 may receive the frequency coefficients converted from the voice / audio signal and obtain the total energy ET of each track structure by the following equation.

는 절대값이 가장 큰 2개의 주파수 계수를 의미한다. 만약, 각 트랙에서 펄스의 수가 4개인 경우, 수학식 1 중

부분은

으로 변경된다.

는 절대값이 가장 큰 4개의 주파수 계수를 의미한다. 이와 같이, 산출부(200)는 펄스의 수에 기초하여 총 에너지를 계산할 수 있다.
here,

Denotes the two frequency coefficients with the largest absolute values. If the number of pulses in each track is four, Equation 1

Part

Is changed to

Denotes the four frequency coefficients with the largest absolute values. As such, the calculator 200 may calculate the total energy based on the number of pulses.

The track structure determiner 210 selects a track structure by comparing the energy concentration or total energy of each track structure calculated by the calculator 200. For example, the track structure determination unit 210 is greater than the EC _structure1 γ × EC _structure2 according to the energy intensity, and selects the select track structure 1, and γ × EC _structure2 is greater than the track frame 2 EC _structure1. Is a value between 0.8 and 1.2. Or, select a γ × ET _structure1 greater than ET _structure2 select a track structure 1, and is larger than the track structure ET _structure1 γ × ET _structure2 2 according to the total energy.

The frequency coefficient allocator 220 allocates the frequency coefficients converted from the voice / audio signal to each track according to the selected track structure. For example, when the track structure determination unit 210 selects the track structure 2, the new coefficient VEC _track (i) is assigned to four tracks as follows.

VEC _tranck1 (i) = spec (4 × i), i = 0, ....., 15

VEC _tranck2 (i) = spec (4 × i + 1), i = 0, ....., 15

VEC _tranck3 (i) = spec (4 × i + 2), i = 0, ....., 15

VEC _tranck4 (i) = spec (4 × i + 3), i = 0, ....., 15

The pulse determiner 230 determines pulses in the order in which the absolute value of the frequency coefficient is large in each track. For example, if two pulses are selected, the frequency coefficient having the largest absolute value and the second largest frequency coefficient are selected in each track.

The quantization unit 240 includes a pulse position quantization unit 242 and a pulse magnitude quantization unit 244. The pulse position quantization unit 242 quantizes the position of the selected pulse, and the pulse magnitude quantization unit 244 quantizes the magnitude of the selected pulse.

The pulse position quantization unit 242 quantizes position information of a pulse selected in each track into bits determined according to the number of positions per track. That is, for example, when the number of positions per track is 8, the position information of the pulse may be quantized to 3 bits. In addition, when the number of positions in the first track is 16, the pulse position information of the track is quantized to 4 bits, and when the number of positions in the second track and the third track is 8, the position information of the pulse selected in the second track and the third track is 3 bits. When the number of positions in the fourth track and the fifth track is four, the position information of the pulse selected in the fourth track and the fifth track may be quantized into 2 bits.

The pulse magnitude quantization unit 244 quantizes the magnitude information of the pulse selected in each track into predetermined bits. For example, when the pulse size quantization unit 244 has two pulses, the pulse size quantization unit 244 may convert the magnitude of the pulse to a log scale to perform vector quantization using a data table previously obtained by an experiment.

The multiplexer 250 multiplexes the pulse position quantization result and the pulse size quantization result quantized by the quantization unit 240 and the track structure determined by the track structure determiner 210 to output a signal having a bit string form.

The demultiplexer 270 demultiplexes the data in the form of a bit string to provide track structure information to the track structure determiner 280, and provides a pulse position quantization value and a pulse magnitude quantization value to the dequantizer 282. .

Inverse quantization unit 282 includes a pulse position inverse quantization unit 284 and a pulse size inverse quantization unit 286. The pulse position dequantization unit 284 dequantizes position information of a predetermined bit quantized by the pulse position quantization unit 242 to restore the pulse position. The pulse size dequantization unit 286 dequantizes the size information of a predetermined bit quantized by the pulse size quantization unit 244 to restore the pulse size.

The pulse generator 288 generates pulses by using the pulse position information quantized by the pulse position dequantizer 284 and the pulse size information quantized by the pulse size dequantizer 286. 290 grasps the frequency coefficient from the pulse generated by the pulse generator 288.

3 is a flowchart of a method of encoding an audio signal, and FIG. 4 is a flowchart of a method of decoding an audio signal, according to an embodiment.

The frequency coefficients converted from the voice / audio signal are received (300) and energy concentration in each track structure is calculated (310).

For example, the track structure includes a sequential structure and an interleave structure. If the number of tracks is 4, the number of pulses in each track is 2, and the frequency coefficient is 64, the energy concentration EC of each track structure can be expressed as follows.

는 절대값이 가장 큰 2개의 주파수 계수를 의미한다. 또한, 총 에너지(ET)는 다음 식에 의해 구해질 수 있다.here,

Denotes the two frequency coefficients with the largest absolute values. In addition, the total energy ET can be obtained by the following equation.

The track structure is determined by comparing the energy concentration or total energy of each calculated track structure (320). For example, the EC _structure1 greater than γ × EC _structure2 according to the energy intensity, selecting the track structure 1 and the γ × EC _structure2 selects a track frame 2 is greater than EC _structure1. Is a value between 0.8 and 1.2. Or, select a γ × ET _structure1 greater than ET _structure2 select a track structure 1, and is larger than the track structure ET _structure1 γ × ET _structure2 2 according to the total energy.

Thereafter, the frequency coefficients converted from the voice / audio signal are allocated to each track according to the selected track structure (330). For example, when track structure 2 is selected, a new coefficient VEC _track (i) is assigned to four tracks as follows.

VEC _tranck1 (i) = spec (4 × i), i = 0, ....., 15

VEC _tranck2 (i) = spec (4 × i + 1), i = 0, ....., 15

VEC _tranck3 (i) = spec (4 × i + 2), i = 0, ....., 15

VEC _tranck4 (i) = spec (4 × i + 3), i = 0, ....., 15

In operation 340, the pulses are determined in order of the absolute value of the frequency coefficient of each track. For example, if two pulses are selected, the frequency coefficient having the largest absolute value and the second largest frequency coefficient are selected in each track.

In operation 350, the position and magnitude information of the selected pulse is quantized. At this time, the position information of the pulse selected in each track is quantized into bits determined according to the number of positions per track. Then, the magnitude information of the pulse selected in each track is quantized into predetermined bits.

Thereafter, the quantized pulse position quantization result, the pulse size quantization result, and the determined track structure information are multiplexed to output a signal having a bit string form (360).

On the other hand, if data in the form of a bit string is input at the time of decoding (400), it is demultiplexed (410). In this case, the demultiplexed result is demultiplexed in the form of bit string to output track structure information, a pulse position quantization value, and a pulse magnitude quantization value.

The track structure is determined based on the track structure information (420), and the pulse position information and the pulse size information are restored by inverse quantization of the pulse position quantization value and the pulse size quantization value (430).

Thereafter, a pulse is generated using the track structure, the pulse position information, and the pulse size information (440), and a frequency coefficient is generated and output from the generated pulse (450).

On the other hand, the above-mentioned audio signal encoding / decoding method can be created by a computer program. The program may also be embodied by being stored in a computer readable media and being read and executed by a computer. The storage medium includes a magnetic recording medium, an optical recording medium and the like.

So far I looked at the preferred embodiments of the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is shown in the claims, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

200: calculator
210, 280: track structure determination unit
220: frequency coefficient allocation unit
230: pulse determination unit
240: quantization unit
250: multiplexer
270 demultiplexer
282: Inverse quantization department
288 pulse generator
290: coefficient generator

Claims (16)

A track structure determiner which determines a track structure using a frequency coefficient converted from the audio signal; And
A frequency coefficient allocator for allocating the converted frequency coefficient to each track according to the track structure determined by the track structure determiner; And
And a quantizer for quantizing at least one pulse in the track based on the assigned frequency coefficient. The method of claim 1,
And a calculator for calculating an energy concentration based on the number of pulses in the track. 3. The method of claim 2,
And the track structure determiner determines a track structure based on the calculated energy concentration. The method of claim 1,
And a calculating unit for calculating the total energy of the pulses based on the number of pulses in the track. 5. The method of claim 4,
And the track structure determiner determines a track structure based on the calculated total energy. The method of claim 1,
And a pulse selector configured to select at least one pulse in the track in order of increasing magnitude of the absolute value of the frequency coefficient by reflecting the absolute value of the assigned frequency coefficient for each track. The method of claim 1, wherein the quantization unit,
A pulse magnitude quantizer for quantizing the magnitude information of the selected pulse; And
And a pulse position quantizer for quantizing position information of the selected pulse. An inverse quantizer for inversely quantizing quantized pulse parameters included in an input bit string and outputting a restored pulse parameter;
A track structure determination unit that determines a track structure based on track parameters included in the input bit stream;
A pulse generator for generating a pulse based on the restored pulse parameter; And
And a coefficient generator for generating frequency coefficients based on the determined track structure and the generated pulses. The method of claim 8, wherein the dequantization unit,
A pulse position dequantizer for restoring pulse position information from the quantized pulse parameter; And
And a pulse size dequantizer for restoring pulse size information from the quantized pulse parameter. Calculating an energy concentration from frequency coefficients converted from the audio signal;
Determining a track structure based on the calculated energy concentration;
Assigning the frequency coefficients to a track based on the determined track structure;
Selecting at least one pulse in a track in the track to which the frequency coefficient is assigned; And
And quantizing the selected pulses. The method of claim 10, after the quantizing step,
And multiplexing the pulse quantization result and the determined track structure information in a bit string form. The method of claim 10, wherein the selecting comprises:
A method of encoding an audio signal, characterized in that some pulses are selected in order of magnitude of absolute magnitude of frequency coefficients. The method of claim 10, wherein the calculating step,
And calculating the energy concentration for each track in consideration of the number of pulses to be selected in the step of selecting the at least one pulse. The method of claim 10, wherein the calculating step,
And calculating the total energy in consideration of the number of pulses to be selected in the step of selecting the at least one pulse. The method of claim 10, wherein the quantizing step comprises:
And quantizing the size information of the selected pulse and the position information of the selected pulse, respectively. Determining a track structure based on the track parameter included in the input bit string;
Inversely quantizing the input bit stream to restore a pulse parameter;
Generating a pulse from the restored pulse parameter; And
And generating and outputting frequency coefficients based on the determined track structure and the generated pulses.

Images