JP2005165056A - Device and method for encoding audio signal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To efficiently generate a bit stream of good sound quality by suppressing a pre-echo while keeping encoding efficiency. <P>SOLUTION: An audio signal encoding device has a frame divider (1) which divides an audio input signal into processing units, an audiopsychology computing element (3) which outputs feature data by the processing units, a block length decision unit (4) which decides whether block lengths are long or short by the processing units based on feature data, a filter bank (2) which calculates permissible error energy by processing units when a block length is long, and puts audio signals of processing units in a block and calculates permissible error energy of each block when block lengths are short, a group decision unit (6) which groups short blocks based on the permissible error energy when the blocks are short, and encoding means (6 to 8) of encoding the audio signal by groups or processing units. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a digital audio signal encoding apparatus and method, and more particularly, to an audio signal encoding apparatus and method using a conversion encoding technique capable of changing a conversion block length.

  In recent years, audio signal coding technology with high sound quality and high efficiency has been applied to DVD-Video audio tracks, portable audio players using semiconductor memory and HDD, music distribution via the Internet, and home servers in home LANs. It is widely used for the storage of music, and is becoming more and more important.

  Many of such audio signal encoding techniques perform time-frequency conversion using a conversion encoding technique. For example, in MPEG-2 AAC and Dolby Digital (AC-3), etc., a filter bank is composed of a single orthogonal transform such as MDCT, and MPEG-1 Audio Layer 3 (MP3) and ATRAC (codes used in MD In the configuration method, a filter bank is configured by connecting subband division filters such as QMF and orthogonal transformation in multiple stages.

  In the transform coding method, basically, the input signals converted into frequency components by the filter bank are grouped into divided frequency bands set based on the human auditory frequency resolution, and each divided frequency band is quantized at the time of quantization. The amount of information is reduced by determining the normalization coefficient and expressing the frequency component by a combination of the normalization coefficient and the quantized spectrum. In MPEG-2 AAC, this divided frequency band is called a scale factor band (SFB), and the normalization coefficient is called a scale factor.

  Furthermore, in these high-efficiency audio coding techniques, by performing masking analysis using human auditory characteristics, by removing spectral components determined to be masked or by allowing masked quantization errors, The amount of information for expressing the spectrum is reduced, and the compression efficiency is increased.

  The masking analysis used in these high-efficiency audio coding techniques is mainly masking by an audible frequency region in silence and frequency masking by a masker in a critical band.

  Since the signal determined to be undetectable by humans by the masking analysis is mainly a signal in the high frequency range, it can be masked even if the quantization error of the high frequency component becomes somewhat large.

  However, in the transform coding method, in the case of a so-called transient state in which there is a sudden change in the audio input signal, the quantization error of the high-frequency component in the part where the sudden change occurs is the signal immediately before or immediately after the sudden change. Therefore, ringing noise occurs.

  As a human auditory characteristic, when a loud sound is generated, it becomes difficult to hear the sound immediately before and immediately after that. This is called a time masking effect. The amount of time that can be heard after a loud sound is relatively long, about 100 msec, although it varies from person to person. However, the masking effect time (pre-masking time) that works immediately before is as short as about 5 to 6 msec. Therefore, when ringing noise is generated, noise before a loud sound is easily detected. This is a phenomenon generally called pre-echo.

  Hereinafter, this phenomenon will be described with reference to the drawings.

  FIG. 11A shows an example of an audio input signal whose amplitude changes abruptly. FIG. 11B shows an example of an audio signal obtained by encoding / decoding this signal with a 2048 sample block that is a normal conversion block length of MPEG-2 AAC. As shown in the drawing, a quantization error in a high frequency region that occurs in a sudden signal change portion affects the entire block.

  As described above, immediately before the portion where the amplitude changes abruptly, a human cannot sense noise due to the time masking effect. However, assuming that the input signal uses the same 44.1KHz sampling frequency as the PCM signal used for music CDs, the block length is converted to time, and the time of 2048 sample blocks is 2048 ÷ 44100 × 1000 = About 46.44 ms, so even if noise is generated in the first half of the time, the pre-masking time will protrude and humans will perceive the pre-echo.

  As a method for suppressing this, in various audio encoding methods, by detecting a sudden change in the input signal and shortening the transform block length, the quantization error of the high frequency component due to the sudden change is changed. The occurrence of pre-echo is suppressed by making it not reach the immediately preceding portion.

  FIG. 12 shows a signal obtained by encoding and decoding the audio signal shown in FIG. 11A with 256 sample blocks having a short block length in MPEG-2 AAC. In this case, the influence of the quantization error in the high frequency region due to the sudden change of the input signal is confined in the 256 sample block in which the change occurs. As before, if this block length is converted to time at the 44.1 KHz sampling frequency, it becomes about 5.80 ms, so that humans can hardly perceive this noise due to the premasking effect, and the pre-echo disappears as a result.

  However, in general, when the block length is shortened, not only does the accuracy of masking analysis decrease due to a decrease in frequency resolution, but also the information consumed by the scale factor because the scale factor band used for quantization increases by the number of blocks. The amount increases, and bits that should be allocated to spectrum information at the time of quantization are consumed by the scale factor, so that the coding efficiency is lowered. As a result, the quantization error cannot be strictly masked particularly at a low bit rate, and noise may be detected more easily than when the block length is long.

  Therefore, in MPEG-2 AAC, when processing with short blocks, multiple blocks are grouped according to the characteristics of the signals included in each block, and the blocks included in the same group share the scale factor, so that The standard defines a mechanism for reducing the bits consumed. This is called grouping.

  In MPEG-2 AAC, by performing appropriate grouping, it is possible to effectively suppress the occurrence of pre-echo while suppressing a decrease in encoding efficiency when conversion is performed with a short block.

  In grouping, since the scale factor is shared by different short blocks, it is originally desirable to group short blocks having similar scale factor patterns into the same group after determining the scale factor. Further, since the scale factor changes according to the change of the input signal, if the group of short blocks does not match the change of the input signal, the quantization error at the time of decoding may increase.

  Also, if too many short blocks that prioritize coding efficiency are grouped together, there is a risk that the quantization error will increase to a level that can be perceptually perceived. In the case of MPEG-2 AAC, according to the standard described in Non-Patent Document 1, a short block is always composed of two or more groups.

  Patent Document 1 discloses a technique for determining a group by calculating a spectrum variation index when adjacent blocks or groups are integrated in all combinations, and comparing the variation index with a threshold value.

  Patent Document 2 proposes a method for sharing the scale factor of adjacent blocks in the context of block floating calculation.

  However, if the frames that become short blocks are continuous, it can be divided into two groups in one frame. However, if the short block frame exists alone, the part before the transient state and the signal It is desirable that at least three or more groups are formed, that is, a portion where the change is drastically changed and a portion where the steady state after the change returns.

JP 2003-108192 A Japanese Unexamined Patent Publication No. 4-304031 ISO / IEC 13818-7

  However, the MPEG-2 AAC standard (see Non-Patent Document 1) describes the format information for storing the grouping information on the bitstream and the decoding method of the grouping information, but determines the group of short blocks. There is no mention of how to do it.

  In addition, the ISO MPEG-4 Ver.1 reference program is implemented so that the grouping pattern is determined in advance and all short block frames are processed with the same group pattern. The input information to be matched does not match the grouping pattern, and the sound quality deteriorates.

  As the simplest group determination method, there is a method of determining a group by determining the degree of similarity of scale factors between adjacent blocks after calculating scale factors of all blocks. However, since the scale factor is actually determined after the quantization process, this method will re-quantize after the group determination, and the processing overhead will be significantly increased. Not realistic.

  Further, in the method described in Patent Document 1, a large amount of calculation is required to calculate the variation index, and it is necessary to perform the calculation repeatedly each time the group integration is determined. As a result, the processing efficiency decreases. . Further, every time iterative calculation is performed, an error accumulated in the variation index increases, and as a result, there is a possibility that grouping that does not match the input signal is performed.

  In the method described in Patent Document 2, only the spectral peak of the block is used and it is determined whether or not the difference between the peaks exceeds a fixed value. However, it is considered where the peak appears on the frequency axis. Not.

  In these prior arts, the minimum number of groups considered from the situation of the previous and subsequent frames is not considered.

  The present invention has been made in view of the above problems, and an object of the present invention is to efficiently create a bit stream with good sound quality by suppressing pre-echo while maintaining encoding efficiency.

  To achieve the above object, an audio signal encoding apparatus according to the present invention divides an audio input signal into processing units, analyzes the audio input signal for each processing unit, and outputs feature data. Analysis means; determination means for determining whether the conversion block length of the audio signal is a long block length or a short block length for each processing unit based on the feature data; and the processing in the case of a long block length. A calculation means for calculating an allowable error energy of a unit, blocking the audio signal of the processing unit in the case of a short block length, and calculating an allowable error energy of each block; and in the case of a short block, based on the allowable error energy Grouping means for grouping short blocks into groups and the conversion block length For each of the groups in the case of heat block, said each processing unit in the case of long blocks, and a coding means for coding the audio signal.

  The audio signal encoding method of the present invention includes a dividing step of dividing an audio input signal into processing units, an analyzing step of analyzing the audio input signal for each processing unit and outputting feature data, and the features A determination step for determining whether the conversion block length of the audio signal is a long block length or a short block length for each processing unit based on the data; and, in the case of a long block length, an allowable error energy of the processing unit In the case of a short block length, the audio signal of the processing unit is blocked, and a calculation step for calculating the allowable error energy of each block; in the case of a short block, the short blocks are grouped based on the allowable error energy. Grouping process to be combined and when the conversion block length is a short block Each serial group, the for each processing unit in the case of long blocks, and a coding step for coding the audio signal.

  If it is determined that the conversion block length of the audio signal of the processing unit to be processed is a short block and the conversion block lengths of the audio signals of the preceding and subsequent processing units are both long blocks, the minimum number of blocks is set to 3. Set. Further, when it is determined that the conversion block length of the audio signal of the processing unit to be processed is a short block and at least one of the conversion block lengths of the audio signals of the preceding and subsequent processing units is a short block, the minimum number of blocks Is set to 2.

  According to the above configuration, it is possible to perform appropriate group determination that matches human auditory characteristics by determining blocks having similar allowable error energies as a reference for allocating code amounts in a frame as the same group, By suppressing the occurrence of pre-echo and preventing a decrease in encoding efficiency due to short block selection, a high-quality bit stream can be efficiently created.

  Furthermore, by using the allowable error energy of the block located immediately before in the threshold used for group determination, it is possible to reliably determine the portion where the input signal is changing as the group division point. It is possible to obtain an accurate grouping result corresponding to a change in signal.

  Also, an appropriate grouping result can be obtained by setting the minimum number of groups based on the block lengths of the preceding and following frames and preventing an increase in quantization error due to many blocks being consolidated into the same group.

  The best mode for carrying out the present invention will be described below in detail with reference to the accompanying drawings.

<First Embodiment>
FIG. 1 is a block diagram illustrating a configuration example of the audio signal encoding apparatus according to the first embodiment.

  In the configuration of FIG. 1, reference numeral 1 denotes a frame divider that divides an audio input signal into frames as processing units. The divided frames are sent to a filter bank 2 and an auditory psychological calculator 3 described later. The auditory psychological arithmetic unit 3 analyzes the input audio input signal in units of frames, calculates an auditory entropy value, and performs masking calculation for each divided frequency band serving as a quantization unit. As a result of this calculation, the auditory entropy (PE) value is output to the block length determiner 4, and the signal-to-mask ratio (Signal Mask Ratio: SMR) for each divided frequency band is output to the group determiner 5.

  The block length determiner 4 compares the PE value sent from the psychoacoustic operator 3 with a predetermined PE threshold value, determines the converted block length, and notifies the filter bank 2 of it. In the first embodiment, the PE threshold is determined in advance and held in the block length determiner 4.

  The filter bank 2 converts the input time signal for each frame input from the frame divider 1 into a frequency spectrum having a block length of a length designated by the block length determiner 4.

  The group determination unit 5 calculates the allowable error energy for each divided frequency band from the SMR value for each divided frequency band transmitted from the psychoacoustic operator 3 and the spectrum string output from the filter bank 2, and the spectrum string is Only in the case of a set of short blocks, the group determination of the short block is performed based on the allowable error energy.

  Reference numeral 6 denotes a bit allocator, which determines the bit amount to be allocated to each divided frequency band with reference to the SMR value for each divided frequency band transmitted from the psychoacoustic calculator 3 and the frequency spectrum output from the filter bank 2. . Reference numeral 7 denotes a quantizer, which calculates a normalization coefficient (scale factor) of the frequency spectrum output from the filter bank 2 for each frequency band and outputs the bit amount allocated to each frequency band output from the bit allocator 6. Quantize the frequency spectrum according to A bit shaper 8 forms a bit stream by appropriately shaping the scale factor and quantized spectrum output from the quantizer 7 into a prescribed format, and outputs the bit stream.

  The audio signal encoding processing operation in the audio signal encoding apparatus having the above configuration will be described below with reference to FIG.

  In the first embodiment, MPEG-2 AAC is described as an example of an encoding method for convenience of explanation, but other encoding methods for performing grouping can be realized by a similar method. Examples of the input audio signal to be encoded include, but are not limited to, audio PCM files and signals obtained by analog / digital conversion of real-time audio signals captured by a microphone.

  First, in step S1, each part shown in FIG. 1 is initialized. At this time, in the first embodiment, 2000 is given as the initial value of the PE threshold value and stored in the block length determiner 4.

  Next, in step S2, it is determined whether or not the input audio signal to be encoded has been completed. If the input signal has been completed, the process proceeds to step S15. If the input signal has not been completed, the process proceeds to step S3. In step S <b> 3, the input audio signal is divided into frames as processing units by the frame divider 1 and sent to the filter bank 2 and the psychoacoustic operator 3. In the case of the MPEG-2 AAC LC (Low-Complexity) profile, one frame is composed of 1024 sample PCM signals. After the frame division, the process proceeds to step S4.

  In step S4, a signal pair mask (SMR) value is calculated by performing auditory entropy (PE) and masking calculation for each frequency band as a quantization unit by the auditory psychological calculator 3 for each frame of the input audio signal. To do. The SMR value is calculated for both the long block length and the short block length. Note that in order to reliably remove aliasing in MDCT, the block length needs to be determined one frame ahead. Therefore, the auditory analysis by the auditory psychological calculator 3 is performed for one frame from the encoding target frame. This is performed for a later frame (hereinafter, a preceding frame). Next, in step S6, the block length determiner 4 compares the calculated PE value of the preceding frame with the PE threshold value preset in the block length determiner 4. Here, if the PE value of the preceding frame is larger than the PE threshold, it is determined that the short block (short block) length is used for the preceding frame, and if not, the long block (long block) length is used for the preceding frame. Judge that. Next, the transform block length of the frame is formally determined based on the determination result of the frame length determined last time and the preceding frame length determined this time. As a result, when the long block length is used for the frame, the process proceeds to step S7. When the short block length is used for the frame, the process proceeds to step S10. As will be described below, the filter bank 2 converts the input signal into a frequency spectrum with a block length according to this determination. The block length determination result of the preceding frame is held in the block length determination unit 4 until it is used for the next block length determination (the preceding frame becomes the encoding target frame).

  In step S10, the filter bank 2 performs orthogonal transform with a long block length on the processing target frame. In the case of MPEG-2 AAC, in order to remove aliasing due to orthogonal transform, a windowing process of the width of the transform block length is performed, and then duplicate transform is performed by MDCT. In the time-frequency conversion, 2048 samples including the processing target frame and the immediately preceding frame are input as one unit, and 1024 frequency spectra are obtained. At this time, when a long block length is used, orthogonal transformation is performed with 2048 samples of the input signal as one block, and 1024 frequency spectra are output. As a result, only one set of spectra divided into 1024 frequency components is obtained. When the process is finished, the process proceeds to step S11.

  In step S11, allowable error energy is calculated from the SMR value of the frame, which is the previous output of the psychoacoustic operator 3, and the spectrum obtained in the filter bank 2 in step S10. Here, the group determination unit 5 obtains an allowable error energy for each divided frequency band for each block in the processing target frame. When the allowable error energy of the divided frequency band b is xmin [b], the energy is calculated by the following equation.

Here, energy [b] is the total energy of the spectrum included in the divided frequency band b. If the i-th spectrum is expressed as xi and the spectrum included in the band b is from the j-th to the k-th, energy [b] is obtained by the following equation.

Further, SMR [b] is an SMR value in the divided frequency band b having a long block length output last time by the auditory psychological calculator 3. Note that the SMR value output from the psychoacoustic operator 3 in step S4 is held in the group determiner 5 until the next processing in which the preceding frame becomes the encoding target frame. This stores both the SMR values for both the long block length and the short block length. When the calculation of the allowable error energy ends, the process proceeds to step S12.
On the other hand, when it is determined in step S6 that the short block length is used for the frame, in step S7, the filter bank 2 performs orthogonal transform on the processing target frame using the short block length. Again, in the case of MPEG-2 AAC, in order to remove aliasing due to orthogonal transform, a windowing process of the width of the transform block length is performed, and then duplicate transform is performed by MDCT. In the time-frequency conversion, 2048 samples including the processing target frame and the immediately preceding frame are input as one unit, and 1024 frequency spectra are obtained. At this time, when the short block length is used, the conversion for outputting 128 frequency spectra with 256 samples of the input signal as one block is performed 8 times while shifting the input signal by 128 samples, and 8 sets of frequency spectra are obtained. Get. When the process is finished, step S9 follows.

  In step S9, the group determination unit 5 calculates the allowable error energy based on the previously stored SMR value of the short block, performs group determination of the short block based on the result, and outputs the result as group information. This group determination is a process of taking a difference in allowable error energy for each divided frequency band between two adjacent blocks and determining a group division point when the sum exceeds a certain threshold. Details of this processing will be described later with reference to FIG. When the group determination ends, the process proceeds to step S12.

  In step S12, the bit assigner 6 assigns bits to each frequency band based on the frequency spectrum output from the filter bank 2, the group information output from the group determiner 5, and the allowable error energy value. Here, bit allocation is performed in two stages. First, the bits to be allocated to the entire frame being processed are determined from the surplus bit amount, the PE value of the frame being processed stored in the bit allocator 6, and the transform block length, and then obtained in step S9 or step S11. Based on the allowable error energy value, the amount of bits to be allocated to each divided frequency band in the frame is determined. Since such processing is common in the transform coding method as in the present invention, detailed description is omitted. Next, the previous frame PE value output from the psychoacoustic operator 3 is stored in the bit allocator 6. When the process is finished, step S13 follows.

  In step S13, the quantizer 7 calculates the scale factor of each frequency band, and quantizes the frequency spectrum according to the bit amount assigned to each frequency band in step S12. When the process is finished, step S14 follows.

  In step S14, the bit shaper 8 shapes the scale factor and quantized spectrum of each frequency band calculated in step S13 into a bit stream according to the format determined by the encoding method, and outputs the bit stream.

  Thereafter, the process returns to step S2, and it is determined whether or not the input audio signal to be encoded has ended. If the input signal has ended, the process proceeds to step S15. In step S15, since the quantized spectrum that has not been output yet due to delay caused by auditory psychological calculation or orthogonal transformation remains, it is shaped into a bit stream and output. When the process is finished, the audio signal encoding process is finished.

  Next, the details of the group determination process performed by the group determination unit 5 in step S9 of FIG. 2 will be described with reference to the flowchart of FIG.

  In step S101, eight sets of allowable error energies are obtained from the previously stored SMR values for the short blocks and the spectrum values of the eight sets of short blocks obtained in step S7. This calculation is the same as the processing when the long block length processing is performed in step S11, and is performed according to Equation 1 and Equation 2. Next, the SMR value of the preceding frame output from the psychoacoustic operator 3 in step S4 is stored in the group determiner 5. In this case, both the SMR value of the long block length and the SMR value of the short block length are stored. When the process is finished, step S102 follows.

  In step S102, the processing target block counter w is reset to zero. The counter w is held in the group determination unit 5. In step S103, it is determined whether the processing target block counter w is 7 or more. If w is less than 7, that is, if group determination between all the short blocks has not been completed, the process proceeds to step S104. When w is 7, that is, when the group determination between all the short blocks is completed, the process proceeds to step S109.

  In step S104, an allowable error energy difference sum S between the wth short block and the w + 1th short block is calculated. In the present embodiment, when each short block is divided into n divided frequency bands, S is obtained by the following equation.

  Here, xmin [w] [b] is an allowable error energy in the divided frequency band b of the block w.

  Next, in step S105, the allowable error energy sum X of the block w is calculated. X is calculated | required by following Formula.

When the process is finished, step S 106 follows.
In step S106, the allowable error energy difference sum S calculated in step S104 and a value obtained by multiplying the allowable error energy sum X of the block w calculated in step S105 by a predetermined coefficient α (0 <α <1). Compare.

  Thus, by using the allowable error energy of the block located immediately before in the threshold used for group determination, it is possible to reliably determine the portion where the input signal is changing as the group division point. Become.

  As a result of the comparison, if the allowable error energy difference sum S is larger (YES in step S106), the process proceeds to step S107, and a determination is made between the block w and the block w + 1 as a group break. Set group boundaries at boundaries. In this process, this group boundary is added to the group information. When the process is finished, step S108 follows.

  On the other hand, if the allowable error energy difference sum S is not larger (NO in step S106), it is determined that the block w and the block w + 1 are in the same group, and the process directly proceeds to step S108. In step S108, the block counter w is incremented, the process returns to step S103, and the above-described processing is repeated.

  If it is determined in step S103 that w is 7, that is, if the group determination between all the short blocks is completed, the process proceeds to step S109, and the allowable error energy is collected for each group according to the determined group information. . In the present embodiment, this process is performed by dividing the total allowable error energy X of the same divided frequency band included in the same group by the number of blocks included in the group. When the process is finished, step S110 follows.

  In step S110, the order of the spectral components is changed according to the determined group information so as to be grouped for each group and each same scale factor. In the present embodiment, the case of MPEG-2 AAC is considered. However, in MPEG-2 AAC, this rearrangement order is determined by the standard and is well known, and thus detailed description thereof is omitted. When the process ends, the group determination process ends, and the process returns to the process of FIG.

  As described above, in the audio signal encoding process according to the first embodiment, blocks having similar allowable error energies serving as a reference for allocating code amounts in a frame are determined to be the same group, so that Appropriate group determination that matches the auditory characteristics is possible, and it is possible to prevent a decrease in encoding efficiency due to short block selection while suppressing the occurrence of pre-echo. As a result, a high-quality bit stream can be efficiently created.

  Furthermore, by using the allowable error energy of the block located immediately before in the threshold used for group determination, it is possible to reliably determine the portion where the input signal is changing as the group division point. It is possible to obtain an accurate grouping result corresponding to a change in signal.

<Second Embodiment>
The present invention can also be implemented as a software program that runs on a general-purpose PC. Hereinafter, this case will be described with reference to the drawings.

  FIG. 4 is a configuration example of an audio signal encoding apparatus using a general-purpose PC in the second embodiment.

  In the configuration shown in the figure, reference numeral 100 denotes a CPU, which performs operations for audio signal encoding processing, logic determination, and the like, and controls each component connected to the bus 102 via the bus 102. A memory 101 stores a basic I / O program, a program code being executed, data necessary for program processing, and the like in the configuration example of the second embodiment. Reference numeral 102 denotes a bus, which transfers an address signal indicating a component to be controlled by the CPU 100, transfers a control signal of each component to be controlled by the CPU 100, and transfers data between the components. Do.

  Reference numeral 103 denotes a terminal for instructing device activation, setting of various conditions and input signals, and encoding start. An external storage device 104 provides a storage area for storing data, programs, and the like. Data, programs, and the like are stored as necessary, and the stored data and programs are called up when necessary.

  Reference numeral 105 denotes a media drive. Programs, data, digital audio signals, and the like recorded on the recording medium are loaded into the audio signal encoding apparatus by the media drive 105 reading. In addition, various data and execution programs stored in the external storage unit 104 can be written in a recording medium.

  A microphone 106 collects sound and converts it into an audio signal. Reference numeral 107 denotes a speaker, which can output arbitrary audio signal data as an actual sound.

  A communication network 108 includes a LAN, a public line, a wireless line, a broadcast wave, and the like. Reference numeral 109 denotes a communication interface, which is connected to a communication network. The audio signal encoding apparatus according to the second embodiment can communicate with an external device via the communication interface 109 via a communication network, and can transmit and receive data and programs.

  The audio signal encoding apparatus according to the second embodiment having the above configuration operates in response to various inputs from the terminal 103. When an input from the terminal 103 is supplied, an interrupt signal is sent to the CPU 100, whereby the CPU 100 reads various control signals stored in the memory 101, and various controls are performed according to the control signals. .

  The audio signal encoding apparatus according to the second embodiment operates when the CPU 100 executes the basic I / O program, the OS, and the audio signal encoding processing program. The basic I / O program is written in the memory 101, and the OS is written in the external storage device 104. When the power of the apparatus is turned on, the OS is read from the external storage unit 104 into the memory 101 by the IPL (Initial Program Loading) function in the basic I / O program, and the operation of the OS is started.

  The audio signal encoding processing program in the second embodiment is a program code based on a flowchart of an audio signal encoding processing procedure shown in FIG.

  FIG. 5 is a content configuration diagram when the audio signal encoding processing program and related data are recorded on a recording medium.

  In the second embodiment, the audio signal encoding processing program and related data are recorded on a recording medium. As shown in the figure, directory information of the recording medium is recorded in the head area of the recording medium, and thereafter, the audio signal encoding processing program which is the content of the recording medium and the audio signal encoding processing related data are stored. It is recorded as a file.

  FIG. 6 is a schematic diagram showing a state in which an audio signal encoding processing program is introduced into the audio signal encoding apparatus of the second embodiment. The audio signal encoding processing program and related data recorded on the recording medium can be loaded into the audio signal encoding apparatus of the second embodiment through the media drive 105 as shown in FIG. When the recording medium 110 is set in the media drive 105, the audio signal encoding processing program and related data are read from the recording medium under the control of the OS and the basic I / O program, and stored in the external storage unit 104. The After that, these information are loaded into the memory 101 at the time of restart and can be operated.

  FIG. 7 shows a memory map in a state where the audio signal encoding device processing program is loaded into the memory 101 and becomes executable.

  At this time, the work area of the memory 101 includes the preceding block length, the current block length, the previous block length, the allowable error energy, the minimum number of groups, the block counter w, the group information PE threshold, the surplus bit amount, the preceding frame SMR, the current frame. The SMR, the previous frame PE value, and the current frame PE value are stored.

  Hereinafter, an audio signal encoding process executed by the CPU 100 in the second embodiment will be described with reference to the flowchart of FIG.

  First, in step S <b> 21, the user designates an input audio signal to be encoded using the terminal 103. In the second embodiment, the audio signal to be encoded may be an audio PCM file stored in the external storage 104, or a signal obtained by analog-digital conversion of a real-time audio signal captured by the microphone 106.

  Next, in step S22, it is determined whether the input audio signal to be encoded has been completed. If the input signal has been completed, the process proceeds to step S35. If the input signal has not been completed, the process proceeds to step S23. In step S23, the input audio signal is divided into frames as processing units for each channel. Similar to the description in the first embodiment, for example, in the case of MPEG-2 AAC, the audio input signal is divided into frames of 1024 samples for each channel. When the process is finished, step S 24 follows.

  In step S24, the psychoacoustic calculation is performed on a frame that is temporally next from the frame to be encoded (hereinafter referred to as the preceding frame). As a result of this calculation, the auditory entropy (PE) of the preceding frame and the SMR value for each divided frequency band, which is also the quantization unit for the preceding frame, are calculated. Here, the SMR value is calculated together with 8 sets for the short block and 1 set for the long block. The calculated PE value is stored in the preceding frame PE value on the memory 101, and all the SMR values are stored in the preceding frame SMR on the memory 101, respectively. When the process is finished, step S25 follows.

  In step S25, the block length of the preceding frame is determined from the result of the auditory analysis performed in step S24. In the second embodiment, this determination is performed by comparing the PE value of the preceding frame with the PE threshold value in the memory 101. That is, when the PE value of the preceding frame is larger than the PE threshold, it is determined as the short block length, and when it is not, it is determined as the long block length. Next, the data stored in the current frame block length on the memory 101 is stored in the previous frame block length in the work area of the memory 101, and further, the data stored in the previous frame block length is stored in the work frame of the memory 101. After storing the current frame block length in the area, the determination result is stored in the preceding frame block length in the work area of the memory 101. Thus, the frame lengths of the current frame and the frames before and after the current frame are stored. When the process is finished, step S26 follows.

  In step S26, the block type of the final current frame is determined from the current frame block length and the preceding frame block length stored on the memory 101. In the case of MPEG-2 AAC, this determination is determined by the method described in the standard document. When the block type is determined, the block length is automatically determined. If the block length is a short block length, the process proceeds to step S27. If the block length is a long block length, the process proceeds to step S30.

  In step S30, based on the determination made in step S26, orthogonal transform based on the long block length is performed on the processing target frame, and in step S31, an allowable error energy for each divided frequency band is calculated. Since the processes performed in steps S30 and S31 are the same as the processes performed in steps S10 and S11 shown in FIG. 2 of the first embodiment, detailed description thereof is omitted here.

  On the other hand, if it is determined in step S26 that the block length is short, orthogonal transform based on the short block length is performed on the processing target frame in step S27. The processing here is the same as the processing executed in step S7 shown in FIG. 2 of the first embodiment. In the case of MPEG-2 AAC, the result is that a set of spectra divided into 128 frequency components is 8 in this case. A pair is obtained. When the process is finished, the process proceeds to step S28.

  In step S28, the minimum number of groups is determined based on the block lengths of frames located before and after the current frame (that is, the preceding frame, the current frame, and the previous frame). That is, when either the previous frame block length on the memory 101 or the preceding frame block length is the short block length, 2 is stored in the minimum number of groups on the memory 101. This is because, in this case, it is sufficient to divide the frame into at least three parts, ie, before and after the transient state and the transient state, together with the previous and subsequent frames, and therefore, it can be grouped into at least two groups according to the MPEG-2 AAC standard. On the other hand, when both the previous and next frames have a long block length, since the frame being processed is a single short block frame, 3 is stored in the minimum number of groups on the memory 101. This is because, as described above, at least three or more groups should be formed: the part before the transient state, the part where the signal changes drastically, and the part that returns to the steady state after the change. . When the process is finished, step S29 follows.

  In step S29, an allowable error energy for each frequency band is calculated from the current frame SMR value of the short block stored on the memory 101 and the eight sets of short block spectra calculated in step S27, and based on the calculated error energy, Determine the group. Details of this process will be described later with reference to FIG. When the process is finished, step S 32 follows.

  In step S32, using the current frame PE value in memory 101, the transform block length, and the allowable error energy obtained in step S29 or step S31, bit allocation is performed in the same procedure as in step S12 of FIG. 2 of the first embodiment. In step S33, the scale factor of each divided frequency band is calculated, the frequency spectrum is quantized according to the bit amount allocated in step S32, and the scale factor and quantized spectrum calculated in step S33 are calculated in step S34. Are formatted according to the format determined by the encoding method and output as a bit stream. In the second embodiment, the bit stream output by this processing may be stored in the external storage 104, or output to an external device connected to the network 108 via the communication interface 109. Also good.

  Step S35 is a process of copying the preceding frame PE value on the memory 101 to the current frame PE value and copying the SMR value stored in the preceding frame SMR to the current frame SMR.

  Thereafter, the process returns to step S22, and when the input signal ends, the process proceeds to step S36. In step S36, quantized spectra that have not yet been output due to delay caused by psychoacoustic computation or orthogonal transformation remain in the memory, so they are shaped into a bit stream and output. When the process is finished, the audio signal encoding process is finished.

  Next, the group determination processing of the second embodiment performed in step S29 of FIG. 8 will be described with reference to the flowchart of FIG. 9, but the same processing as that of FIG. 3 described in the first embodiment will be described. Are given the same reference numerals, and different points from FIG. 3 will be explained.

  Note that the processing target block counter w in FIG. 3 is held on the memory 101.

  In the example shown in FIG. 9, when it is determined that w is 7 in step S103, that is, when the group determination between all short blocks is completed, the process proceeds to step S120, and whether the group determination is correct or not is determined.

  In step S120, it is determined whether or not the obtained number of groups is equal to or greater than the minimum number of groups on the memory 101 determined in step S27. If the number of groups is equal to or greater than the minimum number of groups, the process proceeds to step S109, and the processing described in the first embodiment is performed.

  On the other hand, if the number of groups is less than the minimum number of groups, the process proceeds to step S121 in order to perform group determination again assuming that group determination has failed. In step S121, 0.05 is subtracted from the coefficient α multiplied by the allowable error sum X in step S106, and the process returns to step S102 to redo the group determination. By reducing the coefficient α, the threshold value is lowered in the next determination in step S106, so that the group is divided more finely.

  As described above, according to the second embodiment, audio signal encoding can be performed using a general-purpose PC.

  Further, by performing the processing of step S27 of FIG. 8 and steps S120 and S121 of FIG. 9, it is ensured that an appropriate block determination is performed in consideration of the situation of the frames positioned before and after.

  Needless to say, the above processing can be applied to the processing in the audio signal encoding device of the first embodiment shown in FIG. Further, it is of course possible to execute the processes of FIGS. 2 and 3 described in the first embodiment using a general-purpose PC instead of the processes shown in FIGS.

<Third Embodiment>
In the second embodiment, the example in which the grouping determination is performed based on the sum of the differences of the allowable error energy has been described. However, the grouping determination can be performed based on the peak of the allowable error energy. Hereinafter, the process of performing grouping determination based on the peak of allowable error energy will be described with reference to FIG. The overall flow of the audio signal encoding process is the same as the process shown in the flowchart of FIG. Moreover, it is realizable with the apparatus which has the structure similar to the audio signal encoding apparatus shown in FIG. 4 demonstrated in the said 2nd Embodiment.

  FIG. 10 is a flowchart showing the group determination processing in step S29 of FIG. 8 in the third embodiment.

  Step S201 calculates the allowable error energy for each divided frequency band (SFB) in all the short blocks included in the frame to be processed. The allowable error energy in the third embodiment is also calculated by the same method as the process in step S11 of FIG. 2 described in the first embodiment. When the process is finished, step S 202 follows.

  In step S202, the block counter w on the memory 101 is reset to 1, and in step S203, the SFB position (peak SFB position) at which the allowable error energy in the block 1 reaches a peak is detected. In this process, the SFB position where the allowable error energy is maximized may be obtained.

  Next, in step S204, it is determined whether or not the block counter w is less than 7. If w is less than 7, that is, if all the short blocks have not been determined, the process proceeds to step S205. When w is 7, that is, when the determination between all the short blocks is completed, the process proceeds to step S210.

  In step S205, the peak SFB position of the allowable error energy of the short block w + 1 one block before is detected in the same manner as in step S203. When the process is finished, step S206 follows.

  In step S206, it is determined whether or not the difference between the peak SFB position of the block w and the peak SFB position of the block w + 1 is larger than the threshold value A. In the third embodiment, the threshold value A is determined in advance, and is stored in the work area of the memory 101 when the audio signal encoding device processing program is loaded into the memory 101.

  If the difference between the peak SFB positions is larger than the threshold value A as a result of the determination, it is determined that there is a group break between the block w and the block w + 1, and the process proceeds to step S208. In the group information on the memory 101, the block w and the block w + 1 are determined. Set group boundaries between. When the process is finished, step S209 follows.

  On the other hand, if the difference in the peak SFB position is equal to or smaller than the threshold value A in step S206, the process proceeds to step S207, where the difference between the allowable error energy at the peak SFB position in block w and the allowable error energy at the peak SFB position in block w + 1 is It is determined whether or not the threshold value B is greater. In the third embodiment, it is assumed that the threshold value B is also predetermined and stored in the work area on the memory 101. As a result of the determination, even when the difference in peak allowable error energy is larger than the threshold B, it is determined that the block is between the block w and the block w + 1 as a group break, and in the group information on the memory 101 in step S208, between the block w and the block w + 1. Set the group boundary to. When the process is finished, step S209 follows. If the difference in peak allowable error energy is not greater than the threshold value B, it is determined that the block w and the block w + 1 are in the same group, and the process proceeds directly to step S209.

  In step S209, the group counter w is incremented, the process returns to step S204, and the above-described processing is repeated.

  If it is determined in step S103 that w is 7, that is, if the group determination between all the short blocks is completed, the process proceeds to step S210 to determine whether the group determination is right or wrong.

  In step S210, it is determined whether the obtained number of groups is equal to or greater than the minimum number of groups on the memory 101 determined in step S27 of FIG. If the number of groups is equal to or greater than the minimum number of groups, the process proceeds to step S213, and the processes of steps S109 and S110 of FIG. 3 described in the first embodiment are performed in steps S213 and S214.

  On the other hand, if the number of groups is less than the minimum number of groups, the process proceeds to step S211 to perform group determination again, assuming that group determination has failed. In step S211, 1 is subtracted from the threshold value A on the memory 101, and in step S212, the threshold value B on the memory 101 is appropriately reduced. When the process is finished, the process returns to step S202, and the group determination is performed again.

  By performing this process, it is ensured that an appropriate block determination is performed in consideration of the situation of the frames positioned before and after.

  As described above, it is possible to perform appropriate grouping by handling short blocks with different spectral peak positions as separate groups. In addition, by handling parts with significantly different peak sizes as a separate group, it is possible to reliably handle a part where the input signal changes drastically as a separate group. In this case as well, grouping that matches the change in the input signal is possible. Is possible.

  Further, grouping may be performed by simultaneously detecting a sum of differences of allowable error energy, a peak position difference, and a peak difference, and comprehensively judging the difference.

  In the second embodiment, the recording medium is not particularly mentioned, but this can be applied to any recording medium such as FD, HDD, CD, DVD, MO, and semiconductor memory.

  In addition, the present invention can be implemented with various modifications without departing from the scope of the invention.

It is a block diagram which shows the structural example of the audio signal encoding apparatus in the 1st Embodiment of this invention. It is a flowchart which shows the audio signal encoding process in the 1st Embodiment of this invention. It is a flowchart which shows the group determination process in the 1st Embodiment of this invention. It is a block diagram which shows the structural example of the audio signal encoding apparatus in the 2nd Embodiment of this invention. It is a content block diagram of the storage medium which stored the audio signal encoding process program in the 2nd Embodiment of this invention. It is a schematic diagram which shows a mode that the audio signal encoding process program in the 2nd Embodiment of this invention is introduced. It is a figure which shows the memory map of the state by which the audio signal encoding process program in the 2nd Embodiment of this invention was loaded. It is a flowchart which shows the audio signal encoding process in the 2nd Embodiment of this invention. It is a flowchart which shows the group determination process in the 2nd Embodiment of this invention. It is a flowchart which shows the group determination process in the 3rd Embodiment of this invention. It is explanatory drawing of the concept in the case of encoding the conventional audio signal with a 2048 sample block. It is a figure which shows the concept in the case of encoding the conventional audio signal by a 256 sample block.

Explanation of symbols

1 Frame Divider 2 Filter Bank 3 Auditory Psychological Operation Unit 4 Block Length Determinator 5 Group Determinator 6 Bit Allocation Unit 7 Quantizer 8 Bit Shaper 100 CPU
101 Memory 102 Bus 103 Terminal 104 External Storage Device 105 Media Drive 106 Microphone 107 Speaker 108 Communication Line 109 Communication Interface 110 Recording Medium

Claims (23)

A dividing means for dividing the audio input signal into processing units;
Analyzing means for analyzing the audio input signal for each processing unit and outputting characteristic data;
Determination means for determining whether the conversion block length of the audio signal is a long block length or a short block length for each processing unit based on the feature data;
Calculating means for calculating the permissible error energy of the processing unit in the case of a long block length, blocking the audio signal of the processing unit in the case of a short block length, and calculating the allowable error energy of each block;
In the case of a short block, grouping means for grouping the short blocks into groups based on the allowable error energy;
An audio signal encoding apparatus comprising: encoding means for encoding the audio signal for each group when the transform block length is a short block; and for each processing unit when the transform block length is a long block.   The grouping means determines that the difference group of allowable error energies of consecutive short blocks is the same group when it is smaller than a predetermined ratio of the total allowable error energy of one short block, and if it is larger, The audio signal encoding apparatus according to claim 1, wherein the audio signal encoding apparatus is grouped by determining that there is one.   The grouping unit determines that the group is the same when the difference in the position of the divided frequency band where the allowable error energy of consecutive short blocks is maximum is smaller than a predetermined value, and determines that the group is different when the difference is larger. The audio signal encoding apparatus according to claim 1, wherein the audio signal encoding apparatus is grouped into groups.   In the grouping means, the difference in the position of the divided frequency band where the allowable error energy of the continuous short blocks is maximum is smaller than a predetermined value, and the allowable error energy difference between the short blocks at the position is smaller than the predetermined value. The audio signal encoding apparatus according to claim 1 or 2, wherein the audio signal encoding device is grouped by determining that the groups are the same group, and determining that they are different groups in other cases.   If the determination unit determines that the conversion block length of the audio signal of the processing unit to be processed is a short block and the conversion block lengths of the audio signals of the preceding and subsequent processing units are both long blocks, the minimum number of blocks is set. 5. The audio signal encoding apparatus according to claim 1, further comprising setting means for setting to 3.   The setting means determines that the conversion block length of the audio signal of the processing unit to be processed is a short block and at least one of the conversion block lengths of the audio signals of the preceding and subsequent processing units is a short block by the determination means. 6. The audio signal encoding apparatus according to claim 5, wherein the minimum number of blocks is set to 2.   The audio according to claim 5 or 6, wherein when the number of groups collected by the grouping unit is less than the minimum number of blocks, a grouping determination criterion is changed, and the processing by the grouping unit is re-executed. Signal encoding device.   8. The audio signal encoding apparatus according to claim 1, wherein the feature data is auditory entropy.   9. The audio signal encoding apparatus according to claim 1, wherein the allowable error energy is a product of a reciprocal of a signal-to-mask ratio of each frequency division band and a spectrum energy of each frequency band. .   10. The audio signal encoding apparatus according to claim 1, wherein the encoding format of the encoding means is MPEG-2 / 4 AAC. A dividing step of dividing the audio input signal into processing units;
Analyzing the audio input signal for each processing unit and outputting characteristic data;
A determination step of determining whether the conversion block length of the audio signal is a long block length or a short block length for each processing unit based on the feature data;
Calculating the allowable error energy of the processing unit in the case of a long block length, blocking the audio signal of the processing unit in the case of a short block length, and calculating the allowable error energy of each block;
In the case of a short block, a grouping step of grouping the short blocks into groups based on the allowable error energy;
And a coding step of coding the audio signal for each group when the transform block length is a short block and for each processing unit when the transform block length is a long block.   In the grouping step, when the difference sum of allowable error energies of consecutive short blocks is smaller than a predetermined ratio of the sum of allowable error energies of one short block, it is determined that they are the same group. 12. The audio signal encoding method according to claim 11, wherein the audio signal encoding method is grouped by determining that there is one.   In the grouping step, it is determined that the group is the same group when the difference in position of the divided frequency band where the allowable error energy of consecutive short blocks is maximum is smaller than a predetermined value, and the group is different when the difference is larger. 13. The audio signal encoding method according to claim 11, wherein the audio signal encoding method is grouped into groups.   In the grouping step, when the difference between the positions of the divided frequency bands where the allowable error energy of the continuous short blocks is maximum is smaller than a predetermined value, and the allowable error energy difference between the short blocks at the position is smaller than the predetermined value. The audio signal encoding method according to claim 11 or 12, wherein the audio signal encoding method is grouped by determining that they are in the same group and determining that they are in different groups in other cases.   If it is determined in the determination step that the conversion block length of the audio signal of the processing unit to be processed is a short block and the conversion block lengths of the audio signals of the preceding and subsequent processing units are both long blocks, the minimum number of blocks is set. The audio signal encoding method according to claim 11, further comprising a setting step of setting to 3.   In the setting step, it is determined in the determination step that the conversion block length of the audio signal of the processing unit to be processed is a short block, and at least one of the conversion block lengths of the audio signal of the preceding and subsequent processing units is a short block. 16. The audio signal encoding method according to claim 15, wherein the minimum number of blocks is set to 2.   The audio signal code according to claim 15 or 16, wherein when the number of groups collected in the grouping step is less than the minimum number of blocks, a grouping determination criterion is changed and the grouping step is re-executed. Method.   18. The audio signal encoding method according to claim 11, wherein the feature data is auditory entropy.   19. The audio signal encoding method according to claim 11, wherein the allowable error energy is a product of a reciprocal of a signal-to-mask ratio in each frequency division band and a spectrum energy in each frequency band. .   20. The audio signal encoding method according to claim 11, wherein the encoding format of the encoding step is MPEG-2 / 4 AAC.   A program that can be executed by an information processing apparatus, and that causes the information processing apparatus that has executed the program to function as the audio signal encoding apparatus according to any one of claims 1 to 10.   21. A program executable by an information processing apparatus, comprising program code for implementing the audio signal encoding method according to claim 11.   23. A storage medium readable by an information processing apparatus, wherein the program according to claim 21 or 22 is stored.

Images