JP4822816B2 - Audio signal encoding apparatus and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce a processing amount for quantization, while minimizing sound quality deterioration by not performing auditory mentality analysis, in audio signal encoding which is constituted so that the auditory mentality analysis may be not performed. <P>SOLUTION: A spectrum information amount calculation part 15 calculates a spectrum information amount before quantization. A quantization spectrum information prediction part 16 predicts the spectrum information amount after the quantization based on a frame average bit amount. A quantization step deciding part 7 decides a quantization step of a whole frame by subtracting the spectrum information amount after the quantization from the spectrum information amount before the quantization, and then multiplying the result of the subtraction by a factor obtained from a quantization roughness increment width. A spectrum quantizing part 8 quantizes a frequency spectrum using the quantization step. Then, the spectrum quantizing part 8 performs code amount control, based on a spectrum allocation bit amount which is calculated in a spectrum allocation bit calculation part 12. <P>COPYRIGHT: (C)2007,JPO&amp;INPIT

Description

  The present invention relates to an audio signal encoding apparatus and method.

  In recent years, high-quality and high-efficiency audio signal coding technology has been widely used for DVD-Video audio tracks, portable audio players, music distribution, music storage on home servers in home LAN, etc. The importance is also increasing.

  Many of such audio signal encoding techniques perform time-frequency conversion using a conversion encoding technique. For example, in MPEG-2 AAC, Dolby Digital (AC-3), and the like, a filter bank is configured by a single orthogonal transform such as MDCT (Modified Discrete Cosine Transform). In MPEG-1 Audio Layer III (MP3) and ATRAC (encoding method used for MD (minidisc)), subband division filters such as QMF (Quadrature Mirror Filter) and orthogonal transforms are connected in multiple stages. A filter bank is configured.

  In these transform coding techniques, masking analysis using human auditory characteristics is performed. Then, by removing spectral components determined to be masked or allowing quantization errors to be masked, the amount of information for spectral representation is reduced and compression efficiency is increased.

  In many of these transform coding techniques, the amount of information held in a spectrum is compressed by nonlinearly quantizing the spectrum component. For example, in MP3 and AAC, the amount of information is compressed by raising each spectral component to the power of 0.75.

  In these transform coding techniques, the input signals converted into frequency components by the filter bank are grouped for each divided frequency band set based on the human auditory frequency resolution. Then, the normalization coefficient for each divided frequency band is determined from the auditory analysis result at the time of quantization, and the amount of information is reduced by expressing the frequency component by a combination of the normalization coefficient and the quantized spectrum. This normalization coefficient is actually a variable for adjusting the quantization roughness for each divided band. When the normalization coefficient changes by 1, the quantization roughness changes by one step. In MPEG-2 AAC, this divided frequency band is called a scale factor band (SFB), and the normalization coefficient is called a scale factor.

  In these transform coding systems, the amount of code is controlled by controlling the quantization roughness of the entire frame, which is a coding unit. In many transform coding schemes, the quantization roughness is controlled in steps with an integer power of a certain radix, and this integer is called a quantization step. In the MPEG audio standard, this quantization step for setting the quantization roughness of the entire frame is called “global gain” or “common scale factor”. The scale factor described above is expressed as a relative value to the quantization step, thereby reducing the amount of information necessary for the sign of these variables.

  For example, in MP3 and AAC, when these variables change by 1, the actual quantization roughness changes by 2 3/16 power.

  In the transform coding quantization process, the quantization distortion is controlled such that the quantization error is masked by controlling the scale factor to reflect the result of the auditory operation. At the same time, it is necessary to control the code amount of the entire frame by controlling the quantization step and appropriately adjusting the quantization roughness of the entire frame. Since these two kinds of numerical values that determine the quantization roughness have a significant influence on the encoding quality, it is required to perform these two controls simultaneously and efficiently with caution and accuracy.

  Refer to the MPEG-1 Audio Layer III (MP3) standard (ISO / IEC 11172-3) and the MPEG-2 AAC standard (ISO / IEC 13818-7). It introduces a method for iterative processing using a double loop of a distortion control loop (outer loop) and a code amount control loop (inner loop) as a method for appropriately controlling the scale factor and global gain during quantization. . Hereinafter, this method will be described with reference to the drawings. For convenience, the case of MPEG-2 AAC will be described as an example.

  FIG. 10 shows a simple flowchart of the quantization process described in the ISO / IEC standard.

  First, in step S501, all SFB scale factors and global gains are initialized to 0, and a distortion control loop (outer loop) is entered.

  In the distortion control loop, first, a code amount control loop (inner loop) is executed.

  In the code amount control loop, first, in step S502, one frame, that is, 1024 spectral components are quantized according to the following quantization formula.

In Equation (1), Xq is a quantized spectrum, x _i is a spectrum before quantization (MDCT coefficient), global_gain is a global gain, and scalefac is an SFB scale factor including this spectral component.

  Next, in step S503, the number of used bits for one frame when these quantized spectra are Huffman encoded is calculated and compared with the number of bits assigned to the frame in step S504. If the number of used bits is larger than the allocated number of bits, the global gain is increased by 1 in step S505, the quantization roughness is increased, and the process returns to the spectral quantization in step S502 again. This repetition is performed until the number of bits necessary after quantization becomes smaller than the allocated number of bits, the global gain at this point is determined, and the code amount control loop is terminated.

  In step S506, the quantization error is calculated by dequantizing the spectrum quantized by the code amount control loop and taking the difference from the spectrum before quantization. This quantization error is collected for each SFB.

  In step S507, it is checked whether the scale factor has become larger than 0 in all the SFBs or whether the quantization error is within the allowable error range. If there is an SFB that does not satisfy any of these conditions, the process proceeds to step S508, where the scale factor of the SFB in which the quantization error is not within the allowable error range is increased by 1, and the distortion control loop process is repeated again. The permissible error for each SFB is obtained before the quantization process by auditory calculation.

  As described above, the quantization processing method described in the ISO standard is composed of a double loop, and the global gain and the scale factor can only be controlled in increments of 1. For this reason, the spectral quantization and the bit calculation are repeated many times before the process is converged.

  Here, for example, in the case of MPEG-2 AAC, the spectrum quantization is a process with a large calculation amount because the calculation of the expression (1) is performed 1024 times every time the process is performed once. In addition, since there are 11 types of Huffman code tables that are searched during the bit calculation, if the Huffman code table is fully searched, the calculation amount of the bit calculation inevitably increases.

  Further, in the distortion control loop, the calculation of the quantization error is performed after the inverse quantization, but this processing also has a large amount of calculation. Therefore, it takes a huge amount of processing before the double loop converges.

  In order to solve this problem, various attempts have been made to reduce the processing amount by reducing the number of repetitions of the double loop.

  For example, Patent Document 1 discloses a technique for controlling the common scale factor and the scale factor in a step-by-step manner instead of one step by the number of steps determined according to the characteristics of the Huffman code table. This reduces the number of loops in each double loop and reduces the amount of processing.

  Patent Document 2 discloses a method of executing a normal inner loop after first calculating an estimated value of a quantization step and then calculating a scale factor according to MNR.

  Non-Patent Document 1 discloses a technique for appropriately calculating a scale factor prior to spectrum quantization by using an expression obtained by modifying Expression (1) and an allowable error energy for each SFB obtained by auditory analysis. To do. Thereby, the distortion control loop outside the double loop is removed, and the processing amount is reduced.

  By using these conventional techniques, the convergence of the double loop of the quantization process can be accelerated, and the amount of the quantization process can be reduced to some extent.

  By the way, there is an auditory psychological analysis process as a process which requires a large amount of processing along with the quantization process. Therefore, when processing amount reduction is prioritized over coding efficiency, specifically, for example, when reduction of power consumption is prioritized over sound quality in a relatively inexpensive portable video shooting device, etc. It is also possible to code without any psychoacoustic analysis. At this time, in the quantization process, the scale factor is uniformly set to the same value in all the divided frequency bands, thereby removing the outer distortion control loop and further reducing the processing amount.

JP 2003-271199 A JP 2001-184091 A.D.Duenes, R.Perez, B.Rivas et al., "A robust and efficient implementation of MPEG-2 / 4 AAC Natural Audio Coders", AES 112th Convention Paper (2002)

  However, the conventional technique cannot prevent the double loop described in the ISO standard document from being completely repeated. Therefore, the quantization process cannot be completed unless the spectrum quantization is repeated several to several tens of times, and the amount of the quantization process in the entire encoding process is still large.

  This problem is the same when no psychoacoustic analysis is performed. Even when the scale factor is uniformly set to the same value in all divided frequency bands, only the outer distortion control loop can be omitted, and it is not possible in the prior art to calculate the quantization step before quantization. Is possible. Therefore, in the conventional technique, the spectrum quantization and the bit calculation in the code amount control loop are repeatedly performed, and there is a problem that the processing amount is wasted.

  In addition, when auditory psychological analysis is not performed, PE (auditory entropy), which is the basis of code amount control, is not calculated, so the surplus bits stored in the bit reservoir cannot be assigned to the frame, and sound quality is further deteriorated. The problem of end up occurs.

  Accordingly, an object of the present invention is to reduce the amount of processing required for quantization while minimizing sound quality degradation due to not performing psychoacoustic analysis in audio signal coding configured not to perform psychoacoustic analysis. There is.

Audio signal encoding apparatus according to an aspect of the present invention, an audio input signal dividing means for dividing a frame of the processing unit for each channel, a time signal of 2 consecutive frames obtained from the dividing means into a frequency spectrum Conversion means for performing the conversion process by shifting one frame at a time, spectrum information amount calculation means for calculating the information amount of the frequency spectrum output from the conversion means as the spectrum information amount before quantization, bit rate and sampling based on the frame average bit amount calculated from the rate prediction means for predicting a spectral information amount after quantization, the prediction means from the spectrum information amount before the quantization calculated by the spectral information calculating means The amount of spectral information after quantization predicted in step 1 is subtracted, and the result of subtraction is the quantization roughness. By multiplying a coefficient obtained from the observed width is quantized and determining means for determining before quantization, the frequency spectrum by using the quantization step determined by the determining means quantization step of the entire frame A quantization means ; a bit reservoir for managing an amount of surplus bits in accordance with an encoding standard; a shaping means for shaping a frequency spectrum quantized by the quantization means according to a predetermined format; and the frame average bits, by adding a portion of the excess bit amount accumulated in the bit reservoir and a spectrum allocation bit calculation means for calculating a spectrum assigned bits, the quantizing means, the spectrum calculated by the spectrum allocation bit calculation means and controlling the amount of codes based on the allocated bit amount.

  According to the present invention, in audio signal coding configured not to perform auditory psychological analysis, the amount of processing required for quantization is reduced while minimizing sound quality degradation due to not performing auditory psychological analysis. Can do.

  The present invention is basically based on the idea that the overall quantization roughness can be obtained by dividing the amount of information before quantization by the amount of information after quantization. It is what you want to ask before. Here, since the quantization roughness is generally obtained by multiplying the radix by the quantization step, when taking the logarithm with the base as the base to obtain the quantization step, the division of the information amount is the difference of the information amount. To change. An accurate quantization step can be obtained by multiplying this difference by a coefficient determined by the quantization step size. Further, although the actual amount of information after quantization can only be obtained after quantization, since it can be predicted from the amount of code assigned to the frame, the present invention uses this prediction before quantization. It is an exact quantization step.

  Further, the present invention uses the frame average code amount at the time of prediction before quantization, adds a part of the surplus bit amount accumulated in the bit reservoir at the time of actual quantization, and uses this value as a reference for the code amount. To control. As a result, even if there is some error in the predicted value of the quantization step, the quantization process is completed with a single spectral quantization, and a frame with a large amount of information is automatically left over without auditory analysis. Allow some bits to be allocated.

  In the present invention, after calculating and determining the scale factor for the first time, the quantization step can be calculated almost accurately by the calculation using the value, so that the quantization is performed by almost one spectral quantization and bit calculation. Can be terminated.

  DESCRIPTION OF EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In addition, this invention is not limited to the following embodiment, It shows only the specific example advantageous for implementation of this invention. In addition, not all combinations of features described in the following embodiments are indispensable as means for solving the problems of the present invention.

(First embodiment)
FIG. 1 is a diagram illustrating a configuration of an audio signal encoding device according to the present embodiment. In the figure, a thick line indicates a data signal and a thin line indicates a control signal.

  In the illustrated configuration, the frame divider 1 divides an audio input signal into frames as processing units. The input signal divided into frame units is sent to the filter bank 3. The filter bank 3 performs windowing on the time signal input from the frame divider 1, performs time-frequency conversion with a predetermined block length, and converts the signal into a frequency spectrum.

The spectrum information calculator 15 takes the sum of each frequency spectrum output from the filter bank 3 and calculates the information amount of the frequency spectrum before quantization based on this. The quantization step calculator 7 subtracts the quantized spectral information amount predicted by the quantized spectral information amount predictor 16 described later from the information amount of the spectrum before quantization obtained by the spectral information amount calculator 15. To obtain the quantization step. The spectrum quantizer 8 quantizes each frequency spectrum. The bit shaper 9 shapes the scale factor and the quantized spectrum into a prescribed format as appropriate, creates a bit stream, and outputs it. The bit reservoir 13 manages the number of surplus bits (reserved bits) defined by each coding standard.

  The spectrum allocation bit calculator 12 calculates the number of bits allocated to the quantized spectrum code from the surplus bit amount notified from the bit reservoir 13 and the frame average bit. The quantized spectral information amount predictor 16 performs prediction calculation of the quantized spectral information amount based on the average number of bits assigned to each frame.

  Next, an audio signal encoding operation in the audio signal encoding apparatus having the above configuration will be described. Here, MPEG-2 AAC will be described as an example of an encoding method, but other encoding methods to which a similar quantization method can be applied can be realized by the same method.

  First, prior to processing, each unit is initialized. Initialization sets the quantization step and all scale factor values to zero.

An audio input signal such as an audio PCM signal is divided into frames by the frame divider 1 and sent to the filter bank 3 . In the case of the MPEG-2 AAC LC (Low-Complexity) profile, one frame is composed of PCM signals of 1024 samples, and this signal is transmitted.

  In the filter bank 3, the current input signal for one frame transmitted from the frame divider 1 and the input signal of the previous frame received at the previous conversion are combined for two frames, that is, the time signal of 2048 samples is 1024 samples. Is converted to a frequency component. In the present embodiment, the input signal of the preceding frame is held in a buffer (not shown) in the filter bank 3. The filter bank 3 uses 2048 samples of the input signal as one block, performs windowing, performs MDCT, and outputs 1024 frequency spectra.

  The spectrum information calculator 15 takes the sum of each frequency spectrum output from the filter bank 3 and calculates the information amount of the frequency spectrum before quantization based on this. In the case of MPEG-2 AAC, the information content of the entire spectrum before quantization can be calculated by the following equation.

Here, x _i represents a spectrum before quantization, and the range of i taking the total is one frame, that is, 0 ≦ i ≦ 1023. This is a logarithm with a base of 2 for the sum of each spectrum.

Quantized spectral information amount predictor 1 6 performs prediction calculation of quantized spectral information amount based on the average number of bits allocated to each frame. In this calculation, first, prediction calculation of the total amount of quantized spectrum is performed based on the frame average bit. In the present embodiment, this calculation is performed using an approximate expression created based on the actual measurement of the relationship between the frame bits and the total amount of the quantized spectrum when quantized by a conventional quantizer. For example, if this approximate expression is F (x) and the frame average bit is average_bits, the quantized spectrum prediction total amount can be obtained by the following expression.

However, X _q is a quantized spectrum, and the range of i taking the total is one frame, that is, 0 ≦ i ≦ 1023. In the present embodiment, the frame average bit is calculated in advance from the bit rate, sampling rate, and number of input channels at the time of system initialization. This calculation is well known in the art and will not be described in detail here. The frame average bit held on the system is used while the value calculated at the time of initialization remains unchanged during the encoding process.

  Next, the quantized spectrum total amount is converted into a quantized spectrum information amount. In the present embodiment, this calculation is performed by taking the logarithm of 2 for the total amount of the quantized spectrum obtained by the equation (3). That is, the quantized spectrum information amount is expressed as follows.

  The quantization step calculator 7 subtracts the quantized spectrum information amount output from the quantized spectrum information amount predictor 16 from the information amount of the spectrum before quantization output from the spectrum information amount calculator 15. Thereafter, a quantization step which is the quantization roughness of the entire frame is calculated by multiplying the subtraction result by a coefficient obtained from the step size of the quantization roughness.

  Specifically, in the case of MPEG-2 AAC, the predicted value of the quantization step is obtained by the following equation.

However, X _q is the quantized spectral, x _i is the unquantized spectral, global_gain is global gain (quantization step). The range of i for which the total is taken is one frame, that is, 0 ≦ i ≦ 1023.

  Here, the first term on the right side in the equation (5) is as follows.

  This is the information amount of the entire spectrum before quantization, and is a value calculated by the spectrum information amount calculator 15 according to the equation (2). The second term on the right side is as follows.

  This is the information amount of the spectrum after quantization, and is a value predicted by the quantized spectrum information amount predictor 16 according to the equation (4).

  The equation (5) can be obtained by appropriately modifying the above-described spectrum quantization equation (1) and uniformly substituting 0 for the scale factor scalefac.

  The spectrum allocation bit calculator 12 is notified of the current surplus bit amount managed by the bit reservoir 13 from the bit reservoir 13, for example, adding 20% of these to the frame average bit to make this an allocation bit, and spectral quantization Notify device 8.

  The spectrum quantizer 8 quantizes 1024 frequency spectra according to the quantization step output from the quantization step calculator 7. For example, in the case of MPEG-2 AAC, the quantized spectrum is calculated by the equation (1), and the number of bits consumed in the entire frame is counted.

  Here, when the number of used bits exceeds the number of allocated bits notified from the spectrum allocation bit calculator 12, the quantization step is increased until the number of used bits falls within the spectrum allocation bit number, and the spectrum quantum again. To do. However, the calculation of the quantization step calculator 7 is accurate, and a part of the surplus bit amount is added to the allocated bits in addition to the bit amount when the prediction calculation of the quantization step is performed. For this reason, in many cases, the quantization is completed by performing only one quantization spectrum calculation and bit calculation.

  In addition, a frame in which the amount of used bits is insufficient when spectrum quantization is performed in the quantization step calculated by the quantization step calculator 7 is necessarily a frame whose information amount is originally larger than the average frame. . Therefore, by adding a part of the surplus bits to the assigned bits and performing the spectrum quantization process based on this value, more bits are automatically assigned to such a frame.

  The scale factor and quantized spectrum of each SFB are shaped into a bit stream according to the format determined by the bit shaper 9 and output.

  Finally, the bit shaper 9 notifies the bit reservoir 13 of the actually used bit amount. The bit reservoir 13 calculates the surplus bit amount actually used from the used bit amount notified from the bit shaper 9 and the frame average bit amount, and appropriately adjusts the reserve bits.

  The audio signal encoding apparatus according to the present embodiment described above does not perform auditory psychological analysis with a heavy processing load. In addition, by predicting the amount of spectral information after quantization from the amount of bits allocated to the frame, and using this to calculate the difference in the amount of information that the entire spectrum before and after quantization has, it is possible to quantize before spectral quantization. Predict the conversion step almost accurately. For this reason, the number of repetitions for adjusting the quantization step is reduced, so that the quantization process can be completed quickly. Therefore, the amount of calculation required for the encoding process can be significantly reduced.

  Also, the audio signal encoding apparatus according to the present embodiment predicts the quantization step based on the frame average bit amount, and performs actual spectrum quantization after adding a part of the surplus bit amount uniformly. As a result, even if some prediction error occurs, only one quantization process is required, and reserve bits are automatically allocated to frames with a large amount of original information, so no psychoacoustic analysis is performed. It is possible to minimize the deterioration of sound quality caused by this.

(Second Embodiment)
The present invention can also be implemented as a software program that operates on a general-purpose computer such as a personal computer (PC). Hereinafter, this case will be described with reference to the drawings.

  FIG. 5 is a diagram illustrating a configuration example of the audio signal encoding device according to the present 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 via a bus 102.

  A memory 101 stores a basic I / O program in the configuration example of the present embodiment, a program code being executed, data necessary for program processing, and the like.

  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 an input device such as a keyboard and a mouse, which performs activation of the device, setting of various conditions and input signals, and an instruction to start encoding.

  Reference numeral 104 denotes an external storage device that provides an external storage area for storing data, programs, and the like, and is realized by, for example, a hard disk device. Here, programs and data including the OS are stored, and the stored data and programs are called by the CPU 100 when necessary. As will be described later, an audio signal encoding processing program is also installed in the external storage device 104.

Reference numeral 105 denotes a media drive. Programs, data, digital audio signals, and the like recorded on a recording medium (for example, a CD-ROM) are loaded into the audio signal encoding apparatus by being read by the media drive 105. In addition, various data and execution programs stored in the external storage device 104 can be written in a recording medium. The recording medium is not limited to a CD-ROM, and an HDD, DVD, MO, semiconductor memory, or the like may be used.

  A microphone 106 collects actual 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 the communication network 108. The audio signal encoding apparatus according to this embodiment can communicate with an external device via the communication network 108 via the communication interface 109 to transmit / receive data and programs.

  The audio signal encoding device having such a configuration operates in response to various inputs from the input device 103. When an input from the input device 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 performs various controls according to the control signals. Is called.

In the audio signal encoding apparatus according to the present embodiment, the CPU 100 executes the basic I / O program stored in the memory 101, and loads the OS stored in the external storage device 104 to the memory 101. It works by executing Specifically, when the power of this apparatus is turned on, the OS is read from the external storage device 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

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

  FIG. 6 is a diagram showing a content configuration example when an audio signal encoding processing program and related data are recorded on a recording medium. In the present embodiment, the audio signal encoding processing program and related data are recorded on a recording medium. As shown in the drawing, directory information of the recording medium is recorded in the head area of the recording medium, and thereafter, an audio signal encoding processing program that is the content of the recording medium and audio signal encoding processing related data are files. It is recorded as.

FIG. 7 is a schematic diagram showing the introduction of the audio signal encoding processing program into the audio signal encoding device (PC). The audio signal encoding processing program and related data recorded on the recording medium can be loaded into the apparatus 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 110 and stored in the external storage device 104 under the control of the OS and the basic I / O program. Is done. After that, these information are loaded into the memory 101 at the time of restart and can be operated.

  FIG. 8 is a diagram showing a memory map in a state where the audio signal encoding processing program according to the present embodiment is loaded into the memory 101 and can be executed. As illustrated, the work area of the memory 101 stores, for example, a pre-quantization spectrum aural information amount, a post-quantization spectrum prediction information amount, a spectrum allocation bit, a spectrum buffer, a quantization spectrum, and an input signal buffer. In addition to this, a used bit, a quantization step, a bit rate, a sampling rate, an average allocated bit, and a reserved bit amount are also stored.

  FIG. 9 is a diagram illustrating a configuration example of the input signal buffer in the audio signal encoding device according to the present embodiment. In the configuration shown in the drawing, the buffer size is 1024 × 2 samples, and for convenience of explanation, every 1024 samples are separated by vertical lines. The input signal is input from the right side in increments of 1024 samples for one frame and is processed sequentially from the left side. The bold arrow indicates the flow of the input signal. The illustrated configuration schematically shows an input signal buffer for one channel. In this embodiment, similar buffers are prepared for the channels of the input signal.

  Hereinafter, an audio signal encoding process executed by the CPU 100 in the present embodiment will be described with reference to flowcharts.

  FIG. 2 is a flowchart of audio signal encoding processing in the present embodiment. A program corresponding to this flowchart is included in the audio signal encoding processing program, loaded into the memory 101 as described above, and executed by the CPU 100.

First, step S <b> 1 is a process in which the user designates an input audio signal to be encoded using the terminal 103. In the present embodiment, the audio signal to be encoded may be an audio PCM file stored in the external storage device 104 or a signal obtained by analog / digital conversion of a real-time audio signal captured by the microphone 106. When this process ends, the process proceeds to step S2.

  Step S2 is a process of determining whether or not the input audio signal to be encoded has been completed. If the input signal has ended, the process proceeds to step S11. If not completed, the process proceeds to step S3.

  In step S3, the input signal buffer shown in FIG. 9 shifts the time signal of two frames from the right, that is, 2048 samples to the left by one frame, and newly reads one frame, that is, 1024 samples to the right. This is signal shift processing. This process is performed for all channels included in the input signal. When the process is finished, the process proceeds to step S5.

  In step S5, the time signal of the current frame, that is, the signal of 2048 samples (for two frames) stored in the input signal buffer of FIG. 9 is windowed, and then the time-frequency conversion is performed. As a result, in the case of MPEG-2 AAC, one set of spectra divided into 1024 frequency components is obtained. In this embodiment, all block types are set to a long block length. The calculated 1024 spectra in total are stored in the spectrum buffer in the work area on the memory 101. When step S5 is completed, the process proceeds to step S7.

  Step S7 is a process of calculating the quantization step from the difference between the information amount of the spectrum before quantization and the information amount of the spectrum after quantization. Details of this processing will be described later with reference to FIG. When step S7 is completed, the process proceeds to step S8.

  In step S8, according to the quantization step obtained in step S7, 1024 frequency spectra are quantized and used bits are calculated. Further, only when the number of bits used exceeds the allocated bits stored in the work area on the memory 101, the quantization step is increased and requantization is performed. Details of this processing will be described later with reference to FIG. When step S8 is completed, the process proceeds to step S9.

  Step S9 is a process of shaping the quantized spectrum calculated in step S8 and the scale factor according to a format determined by the encoding method, and outputting the result as a bit stream. In the present embodiment, the bit stream output by this processing may be stored in the external storage device 104, or may be output to an external device connected to the communication network 108 via the communication interface 109. . When step S9 is completed, the process proceeds to step S10.

  Step S10 is a process of correcting the surplus bits stored in the memory 101 from the bit amount used in the bit stream output in step S9 and the frame average bits. When step S10 is completed, the process returns to step S2.

  Step S11 is a process of shaping the quantized spectrum that has not yet been output due to a delay caused by orthogonal transformation or the like in the memory 101, and outputting it after converting it into a bit stream. When this step S11 is completed, the audio signal encoding process ends.

  FIG. 3 is a flowchart showing details of the quantization step prediction process in step S7 described above.

  Step S100 is a process of calculating the amount of information that the spectrum before quantization has. The amount of spectrum information before quantization is obtained by obtaining the total amount of each spectrum component and calculating the logarithm thereof. For example, in the case of MPEG-2 AAC, the amount of spectrum information before quantization can be obtained by the following equation.

  The calculated pre-quantization spectrum information amount is stored in a work area on the memory 101. When step S100 is completed, the process proceeds to step S103.

  Step S103 is a process of performing prediction calculation of the total amount of quantized spectrum using the average number of frames in the memory 101. This prediction calculation is performed by an approximate expression obtained by conducting an experiment in advance. For example, if this approximate expression is F (x) and the frame average bit is average_bits, the quantized spectral prediction total amount can be obtained by the following expression.

  The calculated quantized spectrum prediction total amount is stored in a work area on the memory 101. When step S103 is completed, the process proceeds to step S105.

  Step S105 is a process for calculating the logarithm of the quantized spectrum prediction total obtained in step S103 and calculating the quantized spectrum prediction information amount. For example, MPEG-2 AAC can be calculated by the following equation.

  The quantized spectral information amount calculated by this processing is stored in the work area on the memory 101. When step S105 is completed, the process proceeds to step S108.

  In step S108, a process of subtracting the quantized spectrum prediction information amount obtained in step S105 from the pre-quantization spectrum information amount obtained in step S100 is performed. Next, in step S109, the subtraction result in step S108 is multiplied by a coefficient determined by the step size of the quantization roughness to calculate a global gain, that is, a predicted value of the quantization step. In the case of MPEG-2 AAC, this prediction value is the result of calculating Equation (5) as in the first embodiment.

  The calculated quantization step predicted value is stored in the work area on the memory 101 as a quantization step. The quantization step prediction process is thus completed, and the process returns.

  FIG. 4 is a flowchart showing details of the spectrum quantization process in step S8 described above.

  Step S200 is a process of calculating a spectrum allocation bit by adding a part of the surplus bit amount to the frame average bit stored in the memory 101. For example, in this embodiment, 20% of the surplus bit amount is uniformly added to the frame average bit to obtain the spectrum allocation bit. The calculated spectrum allocation bits are stored in the work area on the memory 101. When step S200 is completed, the process proceeds to step S201.

  Step S201 is a process of quantizing 1024 spectral components stored in the spectral buffer according to the quantization step stored on the memory 101. In the case of MPEG-2 AAC, the quantized spectrum is calculated according to the above equation (1). When step S201 is completed, the process proceeds to step S202.

Step S202 is a process of calculating the number of bits used when encoding all quantized spectrum calculated at step S20 1. For example, in the case of MPEG-2 AAC, since a plurality of quantized spectra are combined and Huffman encoded, the Huffman code table is searched in this process, and the total number of encoded bits is calculated. . The calculated number of used bits is stored in a work area on the memory 101. When step S202 is completed, the process proceeds to step S203.

  Step S203 is a process of comparing the magnitudes of the spectrum allocation bits on the memory 101 and the used bits. As a result of this comparison, if the used bit is larger than the allocated bit, the process proceeds to step S204, the quantization step stored in the memory 101 is increased to reduce the code amount, and the process returns to step S201 again. Quantize the spectrum. However, a substantially accurate quantization step is predicted by the above-described quantization step prediction process (step S7) shown in FIG. 3, and the quantization step is predicted based on the frame average bit. On the other hand, in step S203, since the code amount is controlled based on the spectrum allocation bits obtained by adding a part of the surplus bits to the step S203, step S204 will not be actually executed.

  In addition, as a result of quantization in the predicted quantization step, even if the used bits exceed the frame average bit, if the extra bits do not exceed the added amount, the quantization is completed by one spectrum quantization. Become. Such a frame is originally a frame with a large amount of information, and as a result, more bits are automatically allocated to a frame with a large amount of information.

  If the used bit is smaller than the allocated bit in the comparison in step S203, the spectrum quantization process is terminated and the process returns.

  The audio signal encoding process in the present embodiment described above omits the psychoacoustic analysis process. Then, the amount of information in the spectrum after quantization is predicted from the frame average bit, and the difference from the amount of spectrum information before quantization is calculated, so that the quantization step is almost accurately performed before actual quantization. Predict. Accordingly, it is possible to avoid the adjustment of the quantization step as much as possible without performing the psychoacoustic calculation, and the processing amount for the entire encoding process can be greatly reduced.

  Also, the audio signal encoding apparatus according to the present embodiment predicts the quantization step based on the frame average bit amount, and performs actual spectrum quantization after adding a part of the reserved bit amount uniformly. As a result, even if some prediction error occurs, only one quantization process is required, and reserve bits are automatically assigned to frames with a large amount of original information, so that psychoacoustic analysis is not performed. It is possible to minimize the deterioration of sound quality caused by this.

(Other embodiments)
The present invention can be implemented with various modifications without departing from the scope of the invention.

  For example, although the block switching is not performed at all in the above-described embodiment, the present invention is also applied to an apparatus configured to detect a transient state of an input signal and perform block switching relatively easily without performing auditory analysis. It is possible to apply the invention as well.

  In addition, the present invention may be applied to a system composed of a plurality of devices, or may be applied to an apparatus composed of a single device.

  In the present invention, a program for realizing each function of the above-described embodiments is supplied directly or remotely to a system or apparatus, and a computer included in the system or apparatus reads and executes the supplied program code. Can also be achieved.

  Accordingly, since the functions and processes of the present invention are implemented by a computer, the program code itself installed in the computer also implements the present invention. That is, the computer program itself for realizing the functions and processes is also one aspect of the present invention.

  In this case, the program may be in any form as long as it has a program function, such as an object code, a program executed by an interpreter, or script data supplied to the OS.

  Examples of the recording medium for supplying the program include a flexible disk, hard disk, optical disk, magneto-optical disk, MO, CD-ROM, CD-R, and CD-RW. Examples of the recording medium include a magnetic tape, a non-volatile memory card, a ROM, a DVD (DVD-ROM, DVD-R), and the like.

  The program may be downloaded from a homepage on the Internet using a browser on a client computer. That is, the computer program itself of the present invention or a compressed file including an automatic installation function may be downloaded from a home page to a recording medium such as a hard disk. Further, it is also possible to divide the program code constituting the program of the present invention into a plurality of files and download each file from a different home page. That is, a WWW server that allows a plurality of users to download a program file for realizing the functions and processing of the present invention on a computer may be a constituent requirement of the present invention.

  Further, the program of the present invention may be encrypted and stored in a storage medium such as a CD-ROM and distributed to users. In this case, only users who have cleared the predetermined conditions are allowed to download the key information for decryption from the homepage via the Internet, decrypt the program encrypted with the key information, execute it, and install the program on the computer. May be.

  Further, the functions of the above-described embodiments may be realized by the computer executing the read program. Note that an OS or the like running on the computer may perform part or all of the actual processing based on the instructions of the program. Of course, also in this case, the functions of the above-described embodiments can be realized.

  Furthermore, the program read from the recording medium may be written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer. Based on the instructions of the program, a CPU or the like provided in the function expansion board or function expansion unit may perform part or all of the actual processing. In this way, the functions of the above-described embodiments may be realized.

FIG. 1 is a diagram illustrating a configuration example of an audio signal encoding apparatus according to the first embodiment of the present invention. FIG. 2 is a flowchart of audio signal encoding processing according to the second embodiment of the present invention. FIG. 3 is a flowchart of the quantization step prediction process in the second embodiment of the present invention. FIG. 4 is a flowchart of spectrum quantization processing according to the second embodiment of the present invention. FIG. 5 is a diagram illustrating a configuration example of an audio signal encoding device according to the second embodiment of the present invention. FIG. 6 is a diagram showing a content configuration example of a storage medium storing an audio signal encoding processing program according to the second embodiment of the present invention. FIG. 7 is a schematic diagram showing introduction of an audio signal encoding processing program into a PC according to the second embodiment of the present invention. FIG. 8 is a diagram showing an example of a memory map in the second embodiment of the present invention. FIG. 9 is a diagram illustrating a configuration example of an input signal buffer according to the second embodiment of the present invention. FIG. 10 is a flowchart of the quantization process according to the ISO standard.

Claims (6)

Dividing means for dividing a frame of a processing unit of audio input signals for each channel,
Conversion means for performing a process of converting the time signal of two consecutive frames obtained from the dividing means into a frequency spectrum while shifting each frame by one frame;
A spectrum information amount calculating means for calculating the information amount of the frequency spectrum output from the converting means as a spectrum information amount before quantization;
Prediction means for predicting the amount of spectral information after quantization based on the frame average bit amount calculated from the bit rate and the sampling rate;
The spectral information amount after quantization predicted by the prediction unit is subtracted from the spectral information amount before quantization calculated by the spectral information amount calculation unit , and the subtraction result is obtained from the step size of the quantization roughness. A determination means for determining the quantization step of the entire frame before quantization by multiplying the obtained coefficient;
Quantization means for quantizing the frequency spectrum using the quantization step determined by the determination means ;
A bit reservoir for managing the surplus bit amount according to the encoding standard;
Shaping means for shaping the frequency spectrum quantized by the quantization means according to a predetermined format;
Spectrum allocation bit calculation means for calculating a spectrum allocation bit by adding a part of the surplus bit amount stored in the bit reservoir to the frame average bit;
With
It said quantization means is an audio signal coding apparatus characterized by controlling the code amount based on the spectrum allocation bit amount calculated by the spectral allocation bit calculation means.   The audio signal encoding apparatus according to claim 1, wherein the encoding format is MPEG-1 Audio Layer III.   The audio signal encoding apparatus according to claim 1, wherein the encoding format is MPEG-2 AAC. A dividing step of dividing a frame of a processing unit of audio input signals for each channel,
A conversion step in which the process of converting the time signal of two consecutive frames obtained in the dividing step into a frequency spectrum is performed while shifting one frame at a time;
A spectral information amount calculating step of calculating the information amount of the frequency spectrum obtained in the conversion step as a spectral information amount before quantization;
A prediction step for predicting the amount of spectral information after quantization based on the frame average bit amount calculated from the bit rate and the sampling rate;
The spectral information amount after quantization predicted in the prediction step is subtracted from the spectral information amount before quantization calculated in the spectral information amount calculation step, and the subtraction result is obtained from the step size of the quantization roughness. coefficients obtained by multiplying the a determining step of determining before quantizing the quantization step of the entire frame,
A quantization step of quantizing the frequency spectrum by using the quantization step determined by the determining step,
A shaping step for shaping the frequency spectrum quantized by the quantization step in accordance with a predetermined format,
A spectrum allocation bit calculation step of calculating a spectrum allocation bit by adding a part of the surplus bit amount stored in a bit reservoir that manages the surplus bit amount according to the encoding standard to the frame average bit;
With
The quantization step, the audio signal encoding method characterized by controlling the code amount based on the spectrum allocation bit amount calculated by the spectral allocation bit calculation step.   A program for causing a computer to execute the audio signal encoding method according to claim 4.   A computer-readable storage medium storing the program according to claim 5.

Images