KR101010954B1 - Method for processing audio data, and audio data processing apparatus applying the same - Google Patents

Abstract

An audio data processing method and an audio data processing apparatus using the same are provided. This audio data processing method converts audio data into at least one two-dimensional matrix data and processes the two-dimensional matrix data using a data parallel architecture. As a result, audio data can be processed quickly.

Audio, data parallel architecture, processing

Description

Method for processing audio data, and audio data processing apparatus applying the same

The present invention relates to an audio data processing method and an audio data processing apparatus using the same, and more particularly, to a method for processing audio data using a data parallel architecture and an audio data processing apparatus applying the same.

Data Parallel Architecture refers to a structure that processes data in parallel. When data is processed by the data parallel architecture, a large amount of data can be processed at a time, thereby increasing the processing speed.

Data parallel architectures are commonly used to process two-dimensional images. This is because the processing elements are arranged in the form of a two-dimensional matrix in the processing apparatus using the data parallel architecture, and thus have a structure suitable for processing two-dimensional image data.

However, the video includes audio data as well as image data. Therefore, there is a case in which not only image data but also audio data must be processed together using one processing apparatus.

Accordingly, there is a demand for a method for processing audio data using a data parallel architecture.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to convert audio data into at least one two-dimensional matrix data, and to process the two-dimensional matrix data by using a data parallel architecture. The present invention provides a processing method and an audio data processing apparatus using the same.

According to an embodiment of the present invention, an audio data processing method includes: a conversion step of converting the audio data into two-dimensional matrix data; And a processing step of processing the two-dimensional matrix data in parallel by using a data parallel architecture including a plurality of processing elements arranged in the form of a two-dimensional matrix.

The plurality of processing elements are Y × X and are arranged in a matrix form having X columns and Y rows (X and Y are integers of 1 or more, respectively), and each of the processing elements includes I columns and J rows. Processing a data block containing data in the form of an excitation matrix (I, J each being an integer of 1 or more), wherein the two-dimensional matrix data has X × I columns and Y × J rows, and the processing step includes: X × I × Y × J data included in two-dimensional matrix data may be processed in parallel using the Y × X processing elements.

In the converting step, the [(y-1) × (X × I × J) + (x−1) × (I × J) + (j−1) × I + i] th data of the audio data Mapping to data located in the [(x-1) × I + i] th column and the [(y-1) × J + j] th row of the 2D matrix data, wherein x is 1 An integer value in the range ≤ x ≤ X, wherein y is an integer value in the range 1 ≤ y ≤ Y, i is an integer value in the range 1 ≤ i ≤ I, and j may represent an integer value in the range 1 ≤ j ≤ J. have.

The method may further include inversely converting the processed two-dimensional matrix data into audio data.

On the other hand, the computer-readable recording medium according to an embodiment of the present invention may contain a program for performing the above-described audio data processing method.

On the other hand, according to an embodiment of the present invention, an audio data processing apparatus includes: a converter for converting audio data into two-dimensional matrix data; And a processor configured to process the two-dimensional matrix data in parallel using a data parallel architecture including a plurality of processing elements arranged in a two-dimensional matrix.

The processing unit includes Y × X processing elements arranged in a matrix form having X columns and Y rows (X and Y each being an integer of 1 or more), and each processing element includes I columns and J columns. Processing a data block containing matrix-shaped data having rows (I, J are integers of 1 or more each), and the two-dimensional matrix data has X × I columns and Y × J rows, and the processing unit includes: The X × I × Y × J data included in the two-dimensional matrix data may be processed at once using the Y × X processing elements.

The converting unit may further include [(y-1) × (X × I × J) + (x−1) × (I × J) + (j−1) × I + i] th data among the audio data. Mapping the audio data to the two-dimensional matrix data by mapping the data in the [(x-1) × I + i] -th column and the [(y-1) × J + j] -th row of the two-dimensional matrix data. Where x is an integer value in the range 1≤x≤X, y is an integer value in the range 1≤y≤Y, i is an integer value in the range 1≤i≤I, and j is 1≤j≤J It can also represent an integer value in a range.

The apparatus may further include an inverse transform unit which inversely converts the processed two-dimensional matrix data into audio data.

As described above, according to various embodiments of the present disclosure, an audio data processing method for converting audio data into at least one two-dimensional matrix data and processing the two-dimensional matrix data using a data parallel architecture, and audio data processing using the same It is possible to provide a device so that audio data can be processed using a data parallel architecture.

Hereinafter, with reference to the drawings will be described the present invention in more detail.

1 is a block diagram showing a schematic configuration of an audio device according to an embodiment of the present invention. As shown in FIG. 1, an audio device includes an audio input unit 110, an A / D converter 120, an audio data processing apparatus 200, and an audio output unit 130.

The audio input unit 110 receives an external audio device or a voice of a user. For example, the audio input unit 110 may be implemented to receive a user's voice through a microphone.

The A / D converter 120 converts the input audio signal into a digital audio signal when the analog signal is an analog signal. Therefore, the A / D converter 120 omits the conversion process when the input audio signal is a digital signal.

The audio data processing apparatus 200 processes digital audio data input from the A / D converter 120. For example, the audio data processing apparatus 200 may process noise removal, specific frequency removal, and voice extraction of a user with respect to the input audio data.

In addition, the audio data processing apparatus 200 is implemented using a data parallel architecture. Therefore, since the audio data processing apparatus 200 processes the input audio data in parallel, the audio data processing apparatus 200 may process the audio data at a higher speed.

In general, data parallel architecture is used to process two-dimensional image data. However, the audio data processing apparatus 200 of the present embodiment processes audio data using a data parallel architecture.

To this end, the audio data processing apparatus 200 undergoes a process of converting audio data into two-dimensional matrix data. This is because the two-dimensional image data is configured in the form of two-dimensional matrix data. That is, data values corresponding to each pixel of the 2D image data correspond to the data values of the 2D matrix data.

In this way, when the audio data is converted into two-dimensional matrix data, the data parallel architecture included in the audio data processing apparatus 200 recognizes the audio data as if it is converted into image data and input. Thus, by converting audio data into two-dimensional matrix data (ie, image data form), audio data can be processed using a data parallel architecture.

The structure and function of the audio data processing apparatus 200 will be described later in detail.

The audio output unit 130 outputs the processed audio data through a speaker or the like.

As such, since the audio apparatus includes the audio data processing apparatus 200 using the data parallel architecture, the audio apparatus may simultaneously process a plurality of audio data, thereby processing the audio more quickly. Here, the unit of audio data may be bit or byte, and the number unit of audio data is not limited to the size thereof.

Hereinafter, the audio data processing apparatus 200 will be described in detail with reference to FIG. 2. 2 is a diagram illustrating in detail a structure of an audio data processing apparatus to which a data parallel architecture is applied according to an embodiment of the present invention.

As shown in FIG. 2, the audio data processing apparatus 200 includes a converter 210, a processor 220, and an inverse converter 230.

The converter 210 converts the audio data into at least one two-dimensional matrix data. Specifically, the conversion unit 210 is [(y-1) × (X × I × J) + (x−1) × (I × J) + (j−1) × I + i] th th of audio data. Convert audio data to two-dimensional matrix data by mapping the data to data in the [(x-1) × I + i] th column and the [(y-1) × J + j] th row of the two-dimensional matrix data do. The two-dimensional matrix data has X × I columns and Y × J rows.

Where x is the position of the column of the data block that contains the current data (ie, the position of the column of the processing element that processes the current data), and y is the position of the row of the data block that contains the current data (i.e., processes the current data). Position of the row of the processing element), i denotes the position of the column of the current data within the data block, and j denotes the position of the row of the current data within the data block. That is, x is an integer value in the range 1≤x≤X, y is an integer value in the range 1≤y≤Y, i is an integer value in the range 1≤i≤I, and j is an integer value in the range 1≤j≤J. do.

X denotes the number of columns of the plurality of processor elements PE1 to PE9, Y denotes the number of rows of the plurality of processor elements PE1 to PE9, and I denotes that one processor may process the elements. The number of columns of data blocks, J, represents the number of rows of data blocks that one processing element can process. Here, the data block represents a data unit that can be processed by one processing element, and includes data in a matrix form having J columns and I rows. Two-dimensional matrix data and data blocks will be described in detail later with reference to FIG. 7.

The conversion unit 210 performs the mapping process on all x, y, i, j values, thereby mapping all audio data to two-dimensional matrix data values and converting the audio data into two-dimensional matrix data.

The converter 210 inputs data of each data block to each of the plurality of processing elements PE1 to PE9.

The processor 220 includes Y × X processing elements arranged in a matrix form having X columns and Y rows (X and Y are integers of 1 or more, respectively). In FIG. 2 it is shown to include 3 x 3 processing elements with three columns and three rows. That is, in FIG. 2, the processor 220 may know that Y = 3 and X = 3.

Each processing element processes a data block containing matrix data having J columns and I rows (I and J are integers of 1 or more, respectively). That is, the data block contains a total of J × I data, and each processing element can process J × I data at once. In the present embodiment, it is assumed that J = 3 and I = 3. That is, each processing element can process a total of nine data at once.

According to such a structure, the processor 220 can process X × Y × I × J data at once. For example, in FIG. 2, the processing unit 220 includes nine processing elements PE1 to PE9, and one processing element may process 3 × 3 data blocks, so that a total of 3 × 3 × 3 × 3 = 81 data can be processed at one time.

As described above, the processor 220 has a structure in which a plurality of processing elements are connected in parallel, and such a structure is called a data parallel architecture.

The processing element controller 225 controls the operations of the plurality of processing elements PE1 to PE9. For example, the processing element controller 225 may control whether each processing element is on or off.

As described above, the processor 220 may process the plurality of data in parallel by using the plurality of processing elements PE1 to PE9 and the processing element controller 225.

The inverse transform unit 230 inverts the processed two-dimensional matrix data into audio data. In detail, the inverse transformer 230 may convert the data located in the [(x-1) × I + i] th column and the [(y-1) × J + j] th row of the two-dimensional matrix data among the audio data [( y-1) × (X × I × J) + (x-1) × (I × J) + (j-1) × I + i] inversely converts two-dimensional matrix data into audio data by mapping do.

As such, the audio data processing apparatus 200 may process the audio data using the processor 220 having the data parallel architecture.

Hereinafter, the audio data processing method will be described in detail with reference to FIGS. 3 to 5. 3 is a flowchart provided to explain an audio data processing method according to an embodiment of the present invention.

First, the audio data processing apparatus 200 converts the audio data into at least one two-dimensional matrix data (S310). The conversion process will be described later in detail with reference to FIGS. 4 and 5.

The audio data processing apparatus 200 processes the two-dimensional matrix data using the data parallel architecture (S320). Specifically, the audio data processing apparatus 200 includes Y × X processing elements arranged in a matrix form having X columns and Y rows (X and Y are integers of 1 or more, respectively).

Each processing element then processes a data block containing matrix data having J columns and I rows (I and J are integers of 1 or more, respectively). That is, the data block contains a total of J × I data, and each processing element can process J × I data at once. In this manner, the audio data processing apparatus 200 according to the present embodiment can process X × Y × I × J data at once.

Thereafter, the audio data processing apparatus 200 inversely converts the processed two-dimensional matrix data into audio data (S330). The audio data processing apparatus 200 outputs inversely converted audio data (S340).

Through such a process, the audio data processing apparatus 200 processes audio data using a data parallel architecture.

Hereinafter, a process of converting audio data into 2D matrix data will be described in detail with reference to FIGS. 4 to 7.

4 is a flowchart provided to explain a process of converting audio data into 2D matrix data by the converter 210 according to an embodiment of the present invention.

First, the conversion unit 210 reads X × Y × I × J audio data (S410). In operation S420, the converter 210 extracts a maximum value of the read audio data values. Thereafter, the conversion unit 210 normalizes the audio data values to a value between 0 and 1 using the extracted maximum value (S430). The conversion unit 210 multiplies the normalized audio data values by 255 so that the audio data values are in the range of 0 to 255 (S440).

Next, the conversion unit 210 converts the audio data into two-dimensional matrix data using a conversion algorithm (S450). The conversion algorithm will be described in detail with reference to FIG. 5 below.

5 is a flowchart provided to explain a conversion algorithm, according to an embodiment of the present invention.

In Figure 5, x is the position of the column of the data block containing the current data (i.e., the position of the column of the processing element processing the current data), and y is the position of the row of the data block containing the current data (i.e. The position of the row of the processing element to be processed), i denotes the position of the column of the current data within the data block, and j denotes the position of the row of the current data within the data block. That is, x is an integer value in the range 1≤x≤X, y is an integer value in the range 1≤y≤Y, i is an integer value in the range 1≤i≤I, and j is an integer value in the range 1≤j≤J. do.

First, the y value is initialized to 0 (S510). Then, 1 is added to the y value (S515). Next, the x value is initialized to 0 (S520), and 1 is added to the x value (S525). Thereafter, the j value is initialized to 0 (S530), and 1 is added to the j value (S535). The i value is initialized to 0 (S540), and 1 is added to the i value (S540).

Among the audio data, [(y-1) × (X × I × J) + (x-1) × (I × J) + (j-1) × I + i] th data is replaced with [( x-1) × I + i] maps to the data located in the [(y-1) × J + j] th row (S550).

After that, when i is less than or equal to I (S560-Y), the mapping step S550 is repeated while continuously adding 1 to the i value (S545). If i is greater than I (S560-N), it is checked whether j is less than or equal to J (S570).

Then, steps S535 to S570 are repeated until the j value becomes 1 to J (S560-Y). If j is greater than J (S560-N), it is checked whether x is less than or equal to X (S570).

Then, steps S525 to S580 are repeated until the x value becomes 1 to X (S580-Y). If x is greater than X (S580-N), it is checked whether y is less than or equal to Y (S590).

Then, steps S515 to S590 are repeated until the y value becomes 1 to Y (S590-Y).

Through this process, the X × Y × I × J audio data is mapped to two-dimensional matrix data having X × I columns and Y × J rows. An example in which audio data is converted into two-dimensional matrix data by the above-described algorithm will be described with reference to FIGS. 6 and 7. 6 and 7 illustrate the case where X = 3, Y = 3, I = 3, and J = 3.

6 is a diagram illustrating audio data in graph form according to an embodiment of the present invention. As shown in FIG. 6, audio data is composed of digital values having a specific magnitude at a specific time interval. In FIG. 6, 81 audio data of D1 to D81 are shown.

FIG. 7 illustrates two-dimensional matrix data of nine rows and nine columns according to an embodiment of the present invention. FIG. As shown in FIG. 7, the two-dimensional matrix data includes a first data block 710, a second data block 720, a third data block 730, a fourth data block 740, and a fifth data block ( 750, a sixth data block 760, a seventh data block 770, an eighth data block 780, and a ninth data block 790. In addition, it can be seen that the nine data blocks are arranged in a matrix including three columns (ie, X is 3) and three rows (ie, Y is 3).

In addition, the first data block 710 to the ninth data block 790 are respectively processed by PE1 to PE9 of FIG.

It can be seen that each data block contains nine data in the form of a matrix having three columns (ie, I is 3) and three rows (ie, J is 3). That is, it can be seen that each of the nine processing elements PE1 to PE9 can process nine data at once.

When the 81 audio data D1 to D81 are converted into two-dimensional matrix data by the conversion algorithm described in FIG. 5, the audio data are arranged as two-dimensional matrix data as shown in FIG. Specifically, according to the conversion algorithm of FIG. 5, nine pieces of audio data are allocated to each block first, and in one block, the audio data is sequentially mapped from the upper left to the right. When the top row in the block is full, audio data is sequentially mapped from left to right in the next row. Through this process, two-dimensional matrix data as shown in FIG. 7 is completed.

4, the transform unit 210 inputs the completed two-dimensional matrix data to the data parallel architecture (S460). Specifically, the first data block 710 is in PE1, the second data block 720 is in PE2, the third data block 730 is in PE3, the fourth data block 740 is in PE4, and the fifth data. Block 750 in PE5, sixth data block 760 in PE6, seventh data block 770 in PE7, eighth data block 780 in PE8, ninth data block 790 in PE9 Are input to each.

Through this process, the audio data is converted into two-dimensional matrix data (ie, image data form). And, through this, audio data can be processed using the data parallel architecture.

In this embodiment, there are 81 audio data, 3 x 3 processing elements (i.e., X = 3, Y = 3), and one processing element can process 3 x 3 data (i.e., I = 3, J = 3), but this is just for convenience of description. That is, the number of processing elements and the number of data that can be processed by one processing element can be changed in various ways. Accordingly, the number of columns and rows of two-dimensional matrix data and the number of columns and rows of data blocks vary. .

In this embodiment, an algorithm for converting audio data into two-dimensional matrix data is illustrated in FIG. 5, but is not limited to the algorithm of FIG. 5. In other words, the present invention can be applied to any algorithm as long as the audio data is converted into two-dimensional matrix data. For example, although the algorithm of FIG. 5 places data in the horizontal direction first, an algorithm in which the vertical direction is placed first may be used.

In addition, in the present embodiment, the audio data is described as being converted to the two-dimensional matrix data, but of course, any type of data can be converted to the two-dimensional matrix data if the data in the one-dimensional form of course.

In addition, although the preferred embodiment of the present invention has been shown and described above, the present invention is not limited to the specific embodiments described above, but the technical field to which the invention belongs without departing from the spirit of the invention claimed in the claims. Of course, various modifications can be made by those skilled in the art, and these modifications should not be individually understood from the technical spirit or the prospect of the present invention.

1 is a block diagram showing a schematic configuration of an audio device according to an embodiment of the present invention;

2 is a diagram illustrating in detail the structure of an audio data processing apparatus to which a data parallel architecture is applied according to an embodiment of the present invention;

3 is a flowchart provided to explain an audio data processing method according to an embodiment of the present invention;

4 is a flowchart provided to explain a process of converting audio data into two-dimensional matrix data by a conversion unit according to an embodiment of the present invention;

5 is a flowchart provided to explain a conversion algorithm, according to an embodiment of the present invention;

6 is a diagram illustrating audio data in graph form according to an embodiment of the present invention;

FIG. 7 illustrates two-dimensional matrix data of nine rows and nine columns according to an embodiment of the present invention. FIG.

Explanation of symbols on the main parts of the drawings

110: audio input unit 120: A / D converter

200: audio data processing device 130: audio output unit

210: converter 220: processor

225: processing element controller 230: inverse transform unit

Claims (9)

In the audio data processing method using a plurality of processing elements each consisting of a plurality of data blocks arranged in the form of a first two-dimensional matrix, using a data parallel architecture arranged in the form of a second two-dimensional matrix, Converting the audio data into two-dimensional matrix data arranged in a matrix structure corresponding to the matrix structure of the data parallel architecture; Collectively inputting data corresponding to the row or column of the audio data converted into the two-dimensional matrix data into the row or column of the data parallel architecture; And And processing data corresponding to a row or a column of the input audio data in parallel using the data parallel architecture. The method of claim 1, The plurality of processing elements, Are arranged in a Y × X matrix with X columns and Y rows (X and Y each being an integer of 1 or more), The plurality of data blocks provided in the processing elements, Arranged in the form of a J × I matrix with I columns and J rows, each storing data (I and J are integers of 1 or more, respectively) The data parallel architecture is With X × I columns and Y × J rows, The processing step, And processing the X × I × Y × J data contained in the two-dimensional matrix data in parallel using the Y × X processing elements. The method of claim 2, The conversion step, [(Y-1) × (X × I × J) + (x−1) × (I × J) + (j−1) × I + i] th data of the audio data mapping to data located in the [(x-1) × I + i] th column and the [(y-1) × J + j] th row; X is an integer value in the range 1≤x≤X, y is an integer value in the range 1≤y≤Y, i is an integer value in the range 1≤i≤I, and j is an integer in the range 1≤j≤J And audio data processing. The method of claim 1, Inversely converting the processed two-dimensional matrix data into audio data. A computer-readable recording medium containing a program for performing the audio data processing method according to any one of claims 1 to 4. In the audio data processing apparatus, a plurality of processing elements each consisting of a plurality of data blocks arranged in the form of a first two-dimensional matrix, using a data parallel architecture arranged in the form of a second two-dimensional matrix, A converter for converting the audio data into two-dimensional matrix data arranged in a matrix structure corresponding to the matrix structure of the data parallel architecture; And The data corresponding to the row or column of the audio data converted into the two-dimensional matrix data are collectively input to the row or column of the data parallel architecture, and the data parallel architecture corresponds to the row or column of the audio data to be input. And a processing unit for processing the data to be parallel. The method of claim 6, The plurality of processing elements, Are arranged in a Y × X matrix with X columns and Y rows (X and Y each being an integer of 1 or more), The plurality of data blocks provided in the processing elements, Arranged in the form of a J × I matrix with I columns and J rows, each storing data (I and J are integers of 1 or more, respectively) The data parallel architecture is With X × I columns and Y × J rows, The processing unit, And X X I X Y X J data included in the two-dimensional matrix data in parallel using the Y X X processing elements. The method of claim 7, wherein The conversion unit, [(Y-1) × (X × I × J) + (x−1) × (I × J) + (j-1) × I + i] th data of the audio data is converted into the two-dimensional matrix data. Converting the audio data into the two-dimensional matrix data by mapping to data located in the [(x-1) × I + i] th column and the [(y-1) × J + j] th row of X is an integer value in the range 1≤x≤X, y is an integer value in the range 1≤y≤Y, i is an integer value in the range 1≤i≤I, and j is an integer in the range 1≤j≤J And an audio data processing device. The method of claim 6, And an inverse transform unit which inversely transforms the processed two-dimensional matrix data into audio data.

Images