JP4024784B2 - Audio decoding device - Google Patents

Description

  The present invention relates to an audio decoding apparatus, and more particularly to an oversampling method for decoding audio data encoded in the frequency domain using a time / frequency conversion technique.

  Conventionally, various methods are known for encoding audio signals. As an example, there is a method in which an audio signal is converted from a time domain signal to a frequency domain signal and encoded in the frequency domain. As a method for performing time / frequency conversion, for example, there is a method using a subband filter or MDCT (Modified Discrete Cosine Transform), and MPEG (Moving Picture Image Coding Experts Group) is an encoding method using such a method. Audio.

  In the above-mentioned MPEG audio layer I, in order to efficiently use a psychoacoustic model such as a critical band (a frequency range with a masking effect that a hearing sensitivity decreases at a frequency in the vicinity of a peak of a certain frequency spectrum) Divided into 32 equally spaced frequency widths. Then, the divided signals in each band are subsampled at 1/32 of the original sampling frequency and encoded.

The decoding of the audio data encoded according to the predetermined sampling rate in this way is basically performed by an operation reverse to the above encoding.
FIG. 6 is a block diagram showing the configuration of a conventional MPEG audio decoding apparatus so that the processing flow can be easily understood. In this example, it is assumed that audio data is encoded with a sampling frequency of 44.1 KHz and a bit width of 16 bits.

  In FIG. 6, encoded audio data is first input to the unpack circuit 51. In general, audio data encoded by MPEG audio mainly includes an allocation, a scale factor, and a sample. The unpack circuit 51 separates and extracts the allocation, scale factor, and sample data from the input encoded audio data bit stream.

Each data separated by the unpack circuit 51 is then input to the frequency / time conversion circuit 52. The frequency / time conversion circuit 52, the subband information S _k is a signal in the frequency domain based on the data inputted from the unpack circuit 51 is determined, the sub-band information S in accordance with further below (Equation 1) _A V vector V [i], which is a time domain signal, is _obtained from _k .

  The V vector obtained by the frequency / time conversion circuit 52 is temporarily stored in the V buffer 53 and then given to the filter circuit 54, and is subjected to filter processing using a predetermined filter coefficient. Digital PCM data (44.1 KHz) is generated. The PCM data thus obtained is converted into an analog signal by a 16-bit DAC (digital-analog converter) 55 and output.

  However, in the conventional audio decoding apparatus configured as described above, aliasing noise may occur in the vicinity of half the sampling frequency, and the waveform of the reproduced analog signal may be distorted. . For this reason, when the encoded audio data is decoded and the audio signal before encoding is reproduced, there is a problem that sound reproducibility is deteriorated.

  For example, when a 20 KHz cosine wave is decoded at a sampling rate of 44.1 KHz, the analog audio signal output from the 16-bit DAC 55 of FIG. 6 has a waveform as shown in FIG. Compared with FIG. 10 showing the waveform before encoding, it can be seen that the reproducibility of speech is remarkably deteriorated.

  Conventionally, in order to solve such a problem, a technique in which a 1-bit DAC system 56 as shown in FIG. 7 is used instead of the 16-bit DAC 55 in FIG. 6 has been considered. The 1-bit DAC system 56 includes an interpolator 59 including a FIFO memory 57 and a multiplier / adder 58, and a DAC 60.

  This 1-bit DAC system 56 accumulates PCM data output from the MPEG audio decoder 50 to a certain extent in the FIFO memory 57, and applies digital filter processing to the accumulated PCM data by a multiplier / adder 58. As a result, interpolation data in which data values between discrete real data are estimated is obtained, and analog audio signals are output by performing D / A conversion by the DAC 60 including the interpolation data.

  FIG. 8 is a diagram in which the functional block configuration shown in FIG. 7 is rewritten according to the hardware image. In FIG. 8, the same blocks as those shown in FIG. 7 are denoted by the same reference numerals.

  The multiplier / adder 61 in the MPEG audio decoder 50 shown in FIG. 8 performs frequency / time conversion processing in the frequency / time conversion circuit 52 in FIG. 7 and predetermined filter processing in the filter circuit 54. As various coefficients necessary for performing these processes, those stored in the coefficient ROM / RAM 62 are used.

  The memory 63 shown in FIG. 8 includes a work memory used when performing the frequency / time conversion process and the predetermined filter process, and the V buffer 53 shown in FIG. The PCM data output unit 64 outputs the PCM data generated by the predetermined filter process and stored in the memory 63 to the outside of the MPEG audio decoder 50.

  On the other hand, the coefficient ROM / RAM 65 in the 1-bit DAC system 56 shown in FIG. 8 stores filter coefficients used when the digital adder 58 performs digital filter processing. A plurality of types of filter coefficients are stored, and the sound quality of the reproduced sound is determined to some extent depending on which filter coefficient is used.

  If the 1-bit DAC system 56 as shown in FIG. 7 or FIG. 8 is used, a waveform that is relatively close to the original waveform can be reproduced by using the interpolation data, and deterioration in sound quality can be reduced.

  However, when this 1-bit DAC system 56 is used, in addition to the DAC 60, various configurations such as a multiplier / adder 58, a FIFO memory 57, a coefficient ROM / RAM 65, and the like having considerable calculation capability are required. There is a problem that the circuit scale becomes large and the cost becomes high.

  The present invention has been made to solve such problems, and it is intended to prevent an increase in circuit scale and avoid an increase in cost when a 1-bit DAC system function is provided. Yes.

The audio decoding apparatus of the present invention converts audio data encoded in the frequency domain using time / frequency conversion from information in the frequency domain to information in the time domain, decodes the audio data, and performs interpolation using the decoded audio data. An audio decoding device that generates data and converts digital audio data including the decoded audio data and the interpolated data into an analog audio signal, wherein the time / time for obtaining the decoded audio data is obtained. A multiplier for performing a frequency conversion process and a filter process, and a filter process for generating the interpolation data using the decoded audio data; and a work used when the time / frequency conversion process and the filter process are performed. FI for storing memory and decoded audio data And a memory that functions as O memory, further, performing the time / frequency conversion process, to set the fine processing rate than the processing rate when generating the basic time axis information in accordance with the coding standard Thus, the basic time axis information and the time axis information for interpolating the basic time axis information are generated at the same time.

According to the present invention, audio data encoded in the frequency domain using time / frequency conversion is converted from information in the frequency domain to information in the time domain and decoded, and interpolation data using the decoded audio data is converted. An audio decoding apparatus for generating and converting digital audio data including the decoded audio data and the interpolation data into an analog audio signal, and a multiplier / adder and a memory used for the time / frequency conversion And the multiplier / adder and the memory used for generating the interpolation data are shared by one multiplier / adder and the memory, respectively, so that the interpolation data is obtained during the D / A conversion and the sound reproducibility is obtained. It is possible to realize a configuration for improving the above with a small amount of hardware.
Further, the time / frequency conversion process is performed by setting a processing rate finer than the processing rate in the case of generating basic time axis information in accordance with the encoding standard. And the time axis information for interpolating the basic time axis information are generated at the same time, the time axis information for interpolating the basic time axis information and the decoded audio data are used. Both of the interpolated data can be used.

  Hereinafter, an audio decoding apparatus according to an embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing elemental features of the audio decoding apparatus according to the present embodiment. This audio decoding apparatus is for decoding audio data encoded in the frequency domain using time / frequency conversion. FIG. 1 shows one of a series of decoding processes. Only the part which performs a certain frequency / time conversion process is shown.

  In FIG. 1, (1) (indicated by a circle in the figure) is a frequency / time conversion means, which performs a frequency / time conversion process on the audio data encoded in the frequency domain, thereby encoding standards. Basic time axis information according to the above is generated. For example, when the encoding method is MPEG audio, the frequency / time conversion means (1) uses the V vector V [i which is basic time axis information according to the arithmetic expression based on the standard shown in the above (Expression 1). ] Is generated.

  Further, (2) (displayed with circles in the figure) is an interpolation data generation means, and a frequency / time conversion process performed when basic time axis information is generated by the frequency / time conversion means (1). Interpolation data for interpolating the basic time axis information is generated by the same calculation as the above calculation. For example, when the encoding method is MPEG audio, the interpolation data generation means (2) generates interpolation data V [i] ′ according to (Expression 2) similar to (Expression 1) described below.

  As shown in FIG. 1, the original data used when generating the interpolation data in the interpolation data generation means (2) is used when the basic time axis information is generated in the frequency / time conversion means (1). It is the same as the original data used for.

  (3) (displayed with a circle in the figure) is a multiplex means, and the basic time axis information generated by the frequency / time conversion means (1) and the interpolation data generation means (2). A process of matching with the generated interpolation data is performed. Thereafter, the data output from the multiplex means (3) is subjected to a predetermined process to generate digital decoded audio data. Then, it is converted into an analog audio signal by a D / A conversion means (not shown) and output.

  As described above, according to the embodiment of FIG. 1, in the frequency / time conversion process in the series of decoding processes, in addition to the basic time axis information according to the encoding standard, the basic time axis information is used. Interpolation data for interpolating axis information is generated simultaneously by the same calculation as the calculation of the frequency / time conversion process, so that interpolation data can be obtained without using a complicated 1-bit DAC system. Thus, the reproducibility of voice can be improved by using the interpolation data.

  FIG. 2 is a block diagram showing an example of the configuration of a specific audio decoding apparatus that implements the features of the present invention shown in FIG. FIG. 2 shows an audio decoding apparatus when MPEG audio is used as an example of an encoding method using time / frequency conversion. Audio data is encoded at a sampling frequency of 44.1 KHz. It shall be.

  As shown in FIG. 2, the MPEG audio decoder 1 according to this embodiment includes an unpack circuit 2, a frequency / time conversion circuit 3, a V buffer 4, and a filter circuit 5. The unpack circuit 2 separates allocation, scale factor, and sample data from the input encoded audio data bit stream.

The frequency / time conversion circuit 3 obtains a V vector by performing synthesizer synthesis processing based on each data separated by the unpacking circuit 2. That is, in this synthesizer synthesis process, subband information S _k that is a frequency domain signal is obtained from each data separated by the unpack circuit 2, and further, from the subband information S _k according to (Equation 3) shown below. A V vector V [i], which is a time domain signal, is obtained.

  In (Expression 3), the step width of the sample i is set finer than in the case of the conventional (Expression 1). That is, in (Equation 1), the step width of sample i is 1 as i = 0,1,2,..., Whereas in (Equation 3), i = 0,0.5,1,1.5,2 , ..., the step size of sample i is set to 0.5. As a result, in addition to basic data corresponding to i = 0, 1, 2,..., Interpolation data corresponding to i = 0.5, 1.5,.

  As described above, in this embodiment, interpolation data is not generated at the time of D / A conversion processing after decoding using a 1-bit DAC system, but frequency / time conversion processing performed in a series of decoding processing. At this time, oversampling is performed by calculating with a small step size of the sample, and interpolation data is generated at the same time.

  The V buffer 4 temporarily stores the V vector obtained by the frequency / time conversion circuit 3. The filter circuit 5 generates digital PCM data by subjecting the V vector stored in the V buffer 4 to filter processing using a predetermined filter coefficient. As shown in (Equation 3), in the frequency / time conversion circuit 3, since the processing is performed with the rate set to 1/2 the normal rate, the frequency of the generated PCM data is 88.2KHz. .

  The DAC 6 connected to the subsequent stage of the MPEG audio decoder 1 thus configured converts the digital PCM data output from the MPEG audio decoder 1 into an analog signal and outputs the analog signal. In the present embodiment, since it is not necessary to generate interpolation data at the time of D / A conversion, it is not necessary to use a 1-bit DAC system having a complicated configuration, and a D / A converter having a simple configuration and an inexpensive configuration can be used. It can be used as

  Here, when the audio data obtained by encoding the original cosine waveform shown in FIG. 10 is decoded by the MPEG audio decoder 1 of the present embodiment, the waveform of the analog audio signal output from the DAC 6 is shown in FIG. 4 shows.

  As is clear from the comparison of the waveform of FIG. 4 and the waveform of FIG. 9, according to the present embodiment, a waveform closer to the waveform before encoding shown in FIG. , The reproducibility of voice can be improved. Moreover, in this embodiment, it is possible to improve the reproducibility of voice without using a complicated DAC such as a 1-bit DAC system or providing any other additional configuration.

  In the above embodiment, double the oversampling is realized by finely setting the step width of the sample i to 1/2 of the normal width as shown in (Equation 3), but the step width of the sample i is reduced. If the normal 1 / M is set, M times oversampling can be realized.

  FIG. 5 is a diagram showing a waveform of an analog audio signal output from the DAC 6 when the sample step width is set to 0.125 and 8 times oversampling is executed. As apparent from FIG. 5, compared to the case where the oversampling is performed twice, a waveform closer to the original sound can be reproduced, and the reproducibility of the sound can be further improved.

  As described above, in this embodiment, it is possible to generate more interpolation data by arbitrarily setting the step size of the sample as compared to the case of generating the interpolation data using the 1-bit DAC system. Therefore, there is a merit that the reproducibility of voice can be remarkably improved. In addition, when performing the frequency / time conversion process, the interpolation data can be generated simultaneously according to the same arithmetic expression as the arithmetic expression of the process, so that it is not necessary to change the normal decoding process.

FIG. 3 shows another embodiment of the present invention, and is a diagram showing an example of a hardware configuration of an audio decoding device according to another embodiment.
As shown in FIG. 8, when a 1-bit DAC system is used to generate interpolation data, conventionally, an MPEG audio decoder 50 and a 1-bit DAC system 56 are provided separately.

  On the other hand, in the embodiment shown in FIG. 3, the hardware configuration is simplified by combining the configurations provided redundantly in the MPEG audio decoder 50 and the 1-bit DAC system 56 into one. ing.

  That is, the multiplier / adder 11 in FIG. 3 serves as both the multiplier / adder 61 in the MPEG audio decoder 50 in FIG. 8 and the multiplier / adder 58 in the 1-bit DAC system 56. That is, the multiplier / adder 11 in FIG. 3 includes a synthesizer synthesis process (arithmetic process according to the above (Equation 1)) in the frequency / time conversion circuit 52 in FIG. 7, a predetermined filter process in the filter circuit 54, and a multiplier / adder. The digital filter processing at 58 is performed.

  Also, the coefficient ROM / RAM 12 in FIG. 3 serves as both the coefficient ROM / RAM 62 in the MPEG audio decoder 50 in FIG. 8 and the coefficient ROM / RAM 65 in the 1-bit DAC system 56. That is, various coefficients necessary for frequency / time conversion processing in the frequency / time conversion circuit 52 of FIG. 7, predetermined filter processing in the filter circuit 54, and digital filter processing in the multiplier / adder 58 are stored. .

  Further, the memory 13 in FIG. 3 serves as both the memory 63 in the MPEG audio decoder 50 in FIG. 8 and the FIFO memory 57 in the 1-bit DAC system 56. That is, each process in the multiplier / adder 11 of FIG. 3 described above is performed while using the memory 13 as a work memory.

  The PCM data output unit 14 shown in FIG. 3 outputs the PCM data generated by each process in the multiplier / adder 11 and stored in the memory 13 to the outside. The DAC 15 converts digital PCM data output from the PCM data output unit 14 into an analog signal and outputs the analog signal, and corresponds to the DAC 60 shown in FIG. 7 or FIG.

  As described above, in the embodiment shown in FIG. 3, the conventional MPEG audio decoder 50 and the 1-bit DAC system 56 shown in FIG. , Can reduce the amount of hardware. The total size of the coefficient ROM / RAM memory amount does not change between the case of FIG. 3 and the case of FIG. 8, but the configuration can be simplified by combining them into one memory in FIG.

  In addition, in the embodiment shown in FIG. 3, since it has the function of a 1-bit DAC system, it is possible to generate interpolation data using decoded audio data, and the sound reproducibility is deteriorated. Of course, it can be prevented.

  Still another embodiment may be a combination of the embodiment shown in FIG. 2 and the embodiment shown in FIG. That is, in the present embodiment, the synthesizer combining process performed by the multiplier / adder 11 in the configuration as shown in FIG. 3 is performed according to (Expression 3) instead of (Expression 1).

  In this way, both the interpolation data obtained by finely setting the step size of the sample when performing the frequency / time conversion process and the interpolation data obtained based on the function of the 1-bit DAC system are used. Thus, it is possible to reproduce an analog audio signal, and it can be expected that the reproducibility of sound is further improved with a simple configuration.

  In the embodiment described above, MPEG audio is taken as an example of one of the encoding methods, but if the encoding method adopts the time / frequency conversion method, the frequency / time conversion at the time of decoding is performed. In this case, since the oversampling as described above can be executed, the encoding method is not limited.

  For example, the present invention can also be applied to transform coding systems such as MDCT coding system, subband coding system, AC-3 coding system, or ATRAC (Adaptive TRansform Acoustic Cording).

It is a block diagram which shows an element characteristic. FIG. 2 is a block diagram illustrating a specific configuration example of an audio decoding device that realizes the characteristics illustrated in FIG. 1. It is a block diagram which shows the structural example of the audio decoding apparatus to which this invention is applied. It is a figure which shows the example of the waveform of the analog audio signal obtained when 2 times oversampling is performed in embodiment of FIG. It is a figure which shows the example of the waveform of the analog audio signal obtained when 8 times oversampling is performed in embodiment of FIG. It is a block diagram which shows the structure of the conventional audio decoding apparatus. It is a block diagram which shows the structure at the time of using a 1 bit DAC system in order to solve the conventional problem. It is a block diagram which shows the hardware image of the audio decoding apparatus shown in FIG. It is a figure which shows the example of the waveform of the analog audio signal obtained when a decoding process is performed with the audio decoding apparatus of FIG. It is a figure which shows the example of the waveform of the original audio | voice signal before an encoding.

Explanation of symbols

(1) Frequency / time conversion means (2) Interpolation data generation means (3) Multiplex means 1 MPEG audio decoder 2 Unpack circuit 3 Frequency / time conversion circuit 4 V buffer 5 Filter circuit 6 DAC
11 Multiplier / adder 12 Coefficient ROM / RAM
13 Memory 14 PCM Data Output Unit 15 DAC

Claims (1)

Audio data encoded in the frequency domain using time / frequency conversion is converted from frequency domain information to time domain information and decoded to generate interpolated data using the decoded audio data, and the decoding An audio decoding device configured to convert digital audio data including audio data and the interpolation data into an analog audio signal,
A multiplier for performing the time / frequency conversion process and the filter process for obtaining the decoded audio data, and the filter process for generating the interpolation data using the decoded audio data ;
A work memory used when performing the time / frequency conversion process and the filter process, and a memory functioning as a FIFO memory for storing the decoded audio data ;
Further, the time / frequency conversion process is performed by setting a processing rate finer than the processing rate in the case of generating basic time axis information in accordance with the encoding standard. And an audio decoding device characterized in that the time axis information for interpolating the basic time axis information is generated simultaneously.

Images