JP4301092B2 - Acoustic signal encoding device - Google Patents

Description

  The present invention relates to an acoustic signal encoding apparatus, and more particularly, to an acoustic signal encoding apparatus that encodes an acoustic signal appropriately corresponding to different encoding quality requirements or encoding processing speed requirements.

  In an audio signal encoding device that generates a coded signal by converting information in the frequency domain after converting a digital audio signal in the time domain into the frequency domain, MPEG-1 AUDIO Layer 2 used in video CDs, the Internet, etc. MPEG-1 Layer3 (abbreviated as MP3) used in data distribution in Japan, MPEG-2 AUDIO NBC adopted in BS digital broadcasting, also known as Advanced Audio Coding (abbreviated as AAC), DVD (Digital Versatile) Encoding is performed by an encoding method such as Dolby Digital or MD (minidisc) compression method ATRAC or ATRAC3, which is an audio format of Disk.

  These convert time-domain digital audio signals to frequency domain, according to the characteristics of acoustic signals biased in a specific frequency band, and weighting for each frequency band according to the sensitivity of the auditory considering the characteristics of human hearing, Information compression is performed by reducing or reducing information in a frequency band that is not perceptually important.

  In the above coding method, the part that is greatly related to the quality of hearing is a quantization unit that controls the quantization noise for each frequency band caused by information reduction not to be detected. One of the human auditory characteristics is a masking effect. This masking effect is “when a signal having a relatively high sound pressure exists at a certain frequency, a signal having a low sound pressure near that frequency becomes difficult to hear.” There are two known methods: dynamic masking and static masking (minimum audible limit) that "a signal with a low sound pressure determined by frequency cannot be heard by itself".

  Using these two masking characteristics, the sensitivity depending on the frequency band of the acoustic signal blocked by a predetermined time interval, in other words, the allowable masking level which is a standard that signals below a certain level depending on the frequency band are not detected Can be requested. Psychoacoustic theory states that noise below this allowable masking level cannot be grasped by the general human ear. Therefore, even if the frequency signal is roughly quantized so as to be below the allowable masking level for each predetermined band, sound quality degradation cannot be perceived. The quantization unit is a part that controls the quantization noise below the allowable masking level.

  Quantization noise control is achieved by suppressing the coding amount within the intra-block information allocation amount according to the coding rate and optimizing the quantization noise level along the allowable masking level in units of a predetermined frequency band. That is done. However, the process for controlling the quantization noise occupies most of the encoding process, and is very burdensome.

  In particular, MP3 and AAC incorporate a double repetitive structure for controlling quantization noise, and control the amount of information for each band by adjusting very fine quantization accuracy. One of the iterative processes controls quantization noise for each band to improve the quality of the compressed signal, and the other is for encoding within the allocated information amount for each block according to the compression rate It controls the amount of encoded information. In order to satisfy both conditions, after one repetitive process is completed, the process proceeds to the other repetitive process, and both requirements of quality and encoded information amount are satisfied.

  In recent years, signal processing using not only audio but also video, still images, and other various data has been realized by a single signal processing chip such as a digital signal processor (DSP) and a central processing unit (CPU). It was. However, when a plurality of pieces of information are processed in real time, the system may fail if a delay occurs in the encoding process. Then, errors and omissions occur in the generated encoded signal, and a problem occurs that a correct signal cannot be reproduced on the decoding side, the sound is distorted, the moving image stops, or desired data cannot be acquired.

  Therefore, a method is disclosed in which the processing amount of the acoustic signal encoding is adjusted while monitoring the load of the DSP or CPU to prevent a situation in which real-time processing has failed (see, for example, Patent Documents 1 and 2). ).

  The encoding device described in Patent Document 1 uses CPU load information as an external control signal, and an optimum quantizer from among a plurality of quantizers in the acoustic signal encoding device according to the CPU load amount. Is selected for encoding. The quantizer can be selected for each band, and a heavy load but high efficiency and quality priority type and a light and low efficiency and speed priority type are provided.

  On the other hand, the encoding device described in Patent Document 2 uses the buffer occupancy of the input acoustic signal as an external control signal, and the number of iterations in the quantization means in the acoustic signal encoding device according to the buffer occupancy. The speed is increased by giving an appropriate restriction to the above.

JP 2000-78018 A JP 2001-242895 A

  However, both of the conventional acoustic signal encoding devices described in Patent Documents 1 and 2 are assumed to have abundant input buffer capacity, and a sudden load is applied to the signal processing unit such as a CPU or DSP. In this case, even if the encoded signal of the acoustic signal can be generated, there is a possibility that the signal processing is not in time for other data.

  For example, when encoding a moving image signal and an audio signal in real time, the audio signal must be encoded in a time zone in which the encoding process of the moving image signal is free. The sampling frequency in the audio signal, the frame rate in the video signal, and still image information and text information inserted intermittently are processed at unrelated time intervals, so that a fixed allowable operation time is always guaranteed. is not. As described above, when the amount of load applied to the encoding process is remarkably different, a situation in which the time allowed for the acoustic signal encoding process is extremely short in a certain time zone may occur.

  In addition, an encoding processing apparatus that requires low cost may have a minimum capacity required for encoding or decoding for input / output buffer capacity. Especially in the case of a video signal, the input buffer capacity is very large, and a considerable amount of area must be ensured even when compared with an audio signal. If a delay occurs in the audio signal encoding process, a frame drop (picture drop) occurs in time for the picture signal reserved in the input buffer in the video signal encoding process, and the reproduced video moves momentarily. Gives viewers a sense of incongruity. In particular, since the acoustic signal is focused on maintaining continuity, encoding processing of other information must be sacrificed.

  The above-described conventional high-speed acoustic signal encoding apparatus is effective for a load amount that fluctuates relatively slowly. However, the load on a signal processing unit such as a CPU or DSP that fluctuates rapidly is not effective. Not addressed. In addition, there is a problem in sound quality because the characteristics of the acoustic signal being encoded are not considered and high-speed processing is forcibly performed according to the external control signal even in a time zone that is not suitable for low-quality encoding. There is a case.

  The present invention has been made in view of the above points, and is an acoustic signal code that appropriately responds to different encoding quality requirements or encoding processing speed requirements that change with time, and realizes high sound quality and high-speed processing. An object of the present invention is to provide a device.

In order to achieve the above object, the acoustic signal encoding apparatus of the present invention blocks a digitized acoustic signal at a constant time interval, and then divides the time domain acoustic signal in the block into a plurality of frequency bands. In an acoustic signal encoding apparatus that generates an encoded signal by quantizing an acoustic signal of each frequency band with quantization accuracy corresponding to auditory weighting information for each divided frequency band, digital in the time domain An acoustic characteristic comparison unit that compares the acoustic signal characteristics of the previous block and the current block of the converted acoustic signal or the acoustic signal of each frequency band, determines the correlation between adjacent blocks, and obtains acoustic characteristic comparison information; A first quantizing means for quantizing an acoustic signal in each frequency band of the current block with quantization accuracy according to auditory weighting information for each frequency band; Using the quantization accuracy information we quantize the frequency band of the audio signal, and second quantizing means for performing a simple quantization of the audio signal of each frequency band of the current block is inputted from the outside When the second quantization means is instructed by the encoding control information that specifies the encoding speed to be specified , or when the number of skips in the encoding process is 1 or more, the acoustic signal in each frequency band of the current block Is determined based on the degree of correlation indicated by the acoustic characteristic comparison information, and the second quantum is determined by the quantization determining means based on the acoustic characteristic comparison information. When it is determined that the quantization by the quantizing means is impossible, the non-selecting means for skipping the encoding process of the current block as non-selection of both the first and second quantizing means, and the quantization determining means acoustic When it is determined that the quantization by the second quantizing means is possible based on the sex comparison information, the second quantizing means is selected to quantize the acoustic signal in each frequency band of the block before the current block. Quantization selection execution means to be executed using the quantization accuracy information of the previous block,
When the quantization determination means determines that the quantization by the second quantization means is impossible based on the acoustic characteristic comparison information, the sound of each frequency band of the current block used for the quantization execution by the quantization selection execution means An accumulation means for accumulating the signal and quantization accuracy information corresponding to auditory weighting information for each frequency band, and quantization by the second quantization means based on acoustic characteristic comparison information by the quantization judging means is impossible When it is determined that the coding control information indicates the quantization by the second quantization means after the quantization execution by the quantization selection execution means and the counting means for counting the skip count of the encoding process. When it is detected, the value of the counting means is subtracted by a predetermined value, and the quantizing judging means performs the judgment for the quantization by the second quantizing means of the acoustic signal of each frequency band of the next block, And a coding control means for ending the coding process when it is detected that the coding control information indicates the quantization by the first quantization means after the quantization is executed by the coding selection executing means. It is characterized by.

In this invention, when the quantization control information instructs the quantization by the second quantization means, whether or not the acoustic signal of each frequency band of the current block can be quantized by the second quantization means, When the determination is made based on the degree of correlation indicated by the acoustic characteristic comparison information, and it is determined that the quantization by the second quantizing means is impossible, both the first and second quantizing means are not selected and the current block is not selected. The second quantization of the acoustic signal in each frequency band of the current block from the skipped block at the time when it is determined that the quantization by the second quantization means of the subsequent block is allowed to be performed, from the skipped block Quantization by means can be performed continuously.

  In order to achieve the above object, the acoustic characteristic comparison means calculates a correlation value between the digitized acoustic signal in the time domain of the adjacent previous block and the digitized acoustic signal in the time domain of the current block. Correlation measuring means to be calculated and correlation determining means for determining whether a correlation value measured by the correlation measuring means satisfies a set threshold and generating acoustic characteristic comparison information indicating the degree of correlation between the previous block and the current block It is characterized by providing.

  Further, in order to achieve the above object, the acoustic characteristic comparison means may convert the total energy in the block of the digitized acoustic signal in the time domain or the digitized acoustic signal in the time domain into a plurality of frequency bands. Ratio measurement means for measuring the energy ratio between blocks by comparing the total energy in the block of the acoustic signal for each frequency band obtained by dividing between the previous block and the current block, and according to the energy ratio between blocks Ratio determining means for comparing the value with a preset threshold value and generating acoustic characteristic comparison information indicating the degree of correlation between the previous block and the current block is provided.

  According to the present invention, the first quantization means for performing normal quantization on the acoustic signal of each frequency band of the current block with quantization accuracy according to auditory weighting information for each frequency band; This is selected when the correlation between adjacent blocks is high, and using the quantization accuracy information when quantizing the acoustic signal of each frequency band of the previous block, simplification of the acoustic signal of each frequency band of the current block A second quantizing means for performing quantizing in order to maintain the continuity of the acoustic signal at high load when performing real-time processing with a signal processing chip such as a DSP or CPU that handles multimedia Select one of the first and second quantizing means corresponding to the sudden high-speed processing request and the characteristics of the audio signal, which has prevented the processing system from failing due to frame dropping of the video signal. I did High quality and can realize a high-speed acoustic signal encoding apparatus. In addition, the multimedia encoding processing system can be stabilized by using the acoustic signal encoding apparatus of the present invention.

  Next, the best mode for carrying out the present invention will be described with reference to the drawings. FIG. 1 shows a block diagram of a general acoustic signal encoding apparatus. In the figure, a digital audio signal in the time domain is supplied to the time-frequency conversion unit 1 and is blocked at a certain time interval, and then the digital audio signal in the block is divided into a plurality of frequency bands. Converted to. In addition, the digital audio signal in the time domain is supplied to the frequency band weighting information calculation unit 2 to extract the characteristics of the blocked frequency signal, so that human auditory perception is performed for each of the divided frequency bands. Frequency band weighting information indicating auditory sensitivity using characteristics is calculated.

  Based on the frequency band weighting information output from the frequency band weighting information calculation unit 2, the signal converted into the frequency band output from the time-frequency conversion unit 1 is quantized by the quantization unit 3 for each frequency band. The information is compressed by quantization with the accuracy. The quantized signal obtained by the quantizing unit 3 is supplied to the encoded signal generating unit 4, where it is multiplexed with auxiliary information including frequency band weighting information, individual frequency band quantization accuracy information, etc. A signal encoded signal is generated.

  The frequency band weighting information calculation unit 2 is also referred to as an auditory psychology model, and is a part that calculates the above-described masking level. The masking level is obtained by grouping frequency signals converted into the frequency domain into a plurality of groups depending on individual encoding methods. Since the masking level is determined from the masking amount that each frequency sample exerts on the frequency sample close to itself and the masking amount determined by the minimum audible limit level, the amplitude value of the frequency signal is an important factor. If the blocks have substantially the same level of frequency characteristics, this masking level is also approximately the same between the blocks, and as a result, the same value is applied to the quantization accuracy for quantizing the frequency samples.

  Although the acoustic signal changes from moment to moment, the acoustic signal itself is continuous and often maintains a steady state when viewed in a limited time. FIG. 3 is a graph showing the time transition of an acoustic signal subjected to pulse code modulation (PCM), and its width corresponds to two block lengths (2048 samples) of a general acoustic coding method. The sound source was extracted from a symphony rich in signal changes.

  FIG. 4 shows a frequency characteristic diagram comparing the frequency characteristics in units of blocks using the sound source of FIG. FIG. 4 is a graph showing the frequency characteristics of the first block 1024 samples and the second block 1024 samples in FIG. As shown in FIG. 4, it can be seen that the frequency characteristics are overlapped as if there is a single object, and that the acoustic characteristics between the blocks are very similar. In this case, it can be estimated that the obtained frequency band weighting information is very similar between the preceding and following blocks.

  Therefore, the quantization accuracy for each frequency band applied at the time of quantization is considered to have a very low influence on sound quality even if the information of the former block is applied to the latter block. That is, if the acoustic characteristics of the intra-block acoustic signal are approximated between adjacent blocks, quantization accuracy information at the time of quantization can be inherited from the previous block in the latter block. The present invention focuses on this point and performs low-load quantization as will be described later.

  3 and 4 exemplify the case where the acoustic signal coding method used does not overlap between adjacent blocks on the time axis, but like MDCT (modified discrete cosine transform) like AAC and MP3. It can be easily imagined that the same phenomenon can be seen even in an audio signal encoding method employing frequency conversion that performs a superposed overlap.

  FIG. 2 shows a block diagram of an embodiment of an acoustic signal encoding apparatus according to the present invention. The time-frequency converter 11 and the encoded signal generator 16 of the present embodiment have the same configuration as the time-frequency converter 1 and the encoded signal generator 4 shown in FIG. The calculation unit 14 and the quantization unit 15 are different from the configurations of the frequency band weighting information calculation unit 2 and the quantization unit 3 shown in FIG. 1, and in the present embodiment shown in FIG. It is characterized in that a quantization mode selection unit 13 is added.

  The quantization unit 15 includes a high load quantizer 151, a quantization information accumulator 152, and a low load quantizer 153, and the high load quantizer 151 is selected by a selection signal from the quantization mode selection unit 13. And one of the low load quantizers 153 is selected. The quantization information accumulator 152 accumulates the quantization accuracy information from the high load quantizer 151 and outputs it to the low load quantizer 153.

  Similar to the conventional quantizer, the high load quantizer 151 quantizes the frequency signal of the current block according to the audible weighting information for each frequency band based on the acoustic signal characteristics of the current block. On the other hand, the low-load quantizer 153 performs simple quantization on the frequency signal of the current block, using the quantization accuracy information used when quantizing the previous block as it is.

  In FIG. 2, an input digital audio signal in the time domain is blocked at a fixed time interval by the time-frequency converter 11 and then divided into a plurality of frequency bands for each block. It is converted into a frequency signal consisting of an acoustic signal. The acoustic characteristic comparison unit 12 receives only one of a frequency signal output from the time-frequency conversion unit 11 through a selector in the input unit and a time-domain digital audio signal input to the time-frequency conversion unit 11. It is configured to be input.

  In the case of the configuration shown in FIG. 5 described later, the acoustic characteristic comparison unit 12 determines the correlation between the acoustic characteristics of the current block and the acoustic characteristics of the previous block extracted from the digital acoustic signal in the time domain, and the correlation determination result is used as the acoustic determination result. It transmits to the quantization mode selection part 13 as characteristic comparison information. Note that the acoustic characteristic comparison unit 12 may be configured as shown in FIG. 6 as described later. In this case, the acoustic characteristic comparison unit 12 uses the acoustic signal in the frequency domain from the time-frequency conversion unit 11 to perform acoustic characteristics according to a predetermined comparison target. The comparison information may be acquired. In this case, the detailed characteristic comparison by frequency analysis becomes possible, so that the reliability of the acoustic characteristic comparison information is improved.

  The quantization mode selection unit 13 selects whether to perform a high-speed quantization process or a low-speed quantization process that derives an optimum value in the subsequent quantization unit 15. Assuming a system that processes various types of information in real time, the signal processing of each piece of information must be time-divided and controlled so as not to fail. As the control signal, information (encoding for switching between normal encoding processing or high-speed encoding processing from a signal processing device such as a DSP or CPU including an acoustic signal encoding device to an acoustic signal encoding device) Control information) is input to the quantization mode selection unit 13.

  At this time, if high-speed processing is requested by the encoding control information, the quantization mode selection unit 13 determines whether or not high-speed processing is possible in the current block, so that the acoustic characteristic comparison information from the acoustic characteristic comparison unit 12 To determine whether the quantization accuracy information of the previous block can be applied to the current block.

  If the determination result is true (that is, when the acoustic characteristic comparison information indicates that the acoustic signal characteristics of the current block and the previous block are approximate), the quantization mode selection unit 13 has a low load quantum. The encoder 153 is selected, and if it is false, the encoding process is temporarily interrupted, and the right to occupy the DSP, CPU, or the like is handed over for signal processing used for other information other than the sound signal. The current block performs the encoding process again when the next encoding process request is received. For this purpose, it is necessary to provide a margin for the input buffer for the audio signal. For example, when encoding the video signal at the same time, it is only necessary to secure a much smaller area than increasing the input buffer for the video signal.

  When the encoding control information input from the outside requires normal encoding processing (when it is not a high-speed encoding request), the quantization mode selection unit 13 selects the high-load quantizer 151 that performs conventional optimization. To do.

  Further, the input acoustic signal is supplied to the frequency band weighting information calculation unit 14 in order to perform weighting for each frequency band in the block. Here, the aforementioned masking level is derived, and the initial value or relative value of the quantization accuracy of the frequency samples in the unit frequency band group is determined. Conventionally, frequency band weighting information of each block is derived for each block to be quantized. However, if the low load quantizer 153 is selected in the quantization mode selection unit 13, the low load quantizer 153 performs quantization. Since the frequency signal in the time domain is quantized using the quantization accuracy information of the previous block from the information accumulator 152, the frequency band weighting information of the current block is not necessary.

  Therefore, in this case, the processing of the frequency band weighting information calculation unit 14 can be omitted. The amount of calculation for calculating the frequency band weighting information is the second largest after the quantization unit 15, and omitting this part has a great influence on the speeding up of the encoding process.

  In accordance with the quantization mode information, the quantization unit 15 quantizes the frequency signal using the selected quantizer 151 or 153. Normally, when the encoding mode is specified and the high-load quantizer 151 is selected, the processing for deriving the optimum relationship between the quantization noise level and the encoded information amount is performed as in the conventional acoustic signal encoding device. Optimize. The quantization accuracy information for each band finally obtained is reserved for the quantization information accumulator 152 to be used in the next block.

  Conversely, when the high-speed coding mode is specified and the low-load quantizer 153 is selected, the frequency signal is immediately quantized using the quantization accuracy information of the previous block stored in the quantization information storage 152. To do. In general, the processing performed by the low-load quantizer 153 in the quantizing unit 15 with the highest load is only the quantization of frequency samples and the limitation of the amount of allocated information, and can be completed in a very short time. . As the quantization accuracy information of the previous block used when the low-load quantizer 153 performs low-load quantization on the frequency signal of the current block, weighting information for each frequency band of the previous block is independently used. Or in combination with quantization accuracy information.

  In order to satisfy the allocated information amount, it is sufficient to limit the bandwidth. If the information is allocated by quantizing from the low frequency band and the allocated information amount is exceeded, the allocation information amount requirement can be satisfied if the higher band is not encoded. In general, the acoustic signal is gradually attenuated. If the acoustic characteristic comparison unit 12 detects and determines the attenuation tendency of the acoustic signal, the current block uses the same quantization accuracy as compared to the previous block. Even if encoding is performed, the amount of codes does not increase significantly, and the same frequency band as that of the previous block can be maintained.

  Finally, the quantized value (quantized signal) of the frequency signal output from the quantizing unit 151 is multiplexed with auxiliary information accompanying the encoding process and the encoded signal generating unit 16 and generated as an encoded signal. Is output.

  Next, the configuration of the acoustic characteristic comparison unit 12 will be described in more detail. First, the structure which determines the correlation between adjacent blocks using the acoustic signal of a time domain is demonstrated with FIG. FIG. 5 is a block diagram of the acoustic characteristic comparison unit 12a that performs time shift to achieve correlation while obtaining autocorrelation. In FIG. 5, the input time domain acoustic signal is stored in the previous block acoustic signal buffer 121, and a phase shift modifier for correcting the phase shift of the main frequency component in the unit length block depending on the encoding method. 122, and is corrected so that the correlation with the previous block sound signal stored in the previous block sound signal buffer 121 can be obtained more accurately.

  The correlation measuring unit 123 is configured to obtain a difference between the acoustic signal of the previous block from the previous block acoustic signal buffer 121 and the acoustic signal of the current block in which the phase shift is corrected by the phase shift modifier 122 and the phase is aligned with the previous block. Autocorrelation is taken to determine the autocorrelation between the adjacent previous block and the current block. The correlation determining unit 124 determines whether the correlation value obtained by the correlation measuring unit 123 satisfies a certain predetermined threshold value. If the correlation value is satisfied, the low load quantum using the quantization accuracy value of the previous block is determined. The quantization mode selection unit 13 is designated by the acoustic characteristic comparison information so that the quantizer 153 is used.

  As another example, the degree of correlation between blocks can be measured using an energy ratio or an amplitude ratio. For example, the total energy in the blocks may be compared using an acoustic signal in the time domain or the frequency domain.

  The quantization accuracy is determined from the MNR (masking level to noise level ratio) obtained by the band weighting information calculation unit based on the frequency signal. Therefore, the correlation between the blocks can be measured more accurately by determining the frequency signal than the time domain signal. Specifically, it can be determined from the block average of the amplitude ratio or energy ratio of the samples having the same frequency between blocks, the standard deviation value indicating the degree of concentration, and the like. Furthermore, calculation is performed using the group representative, maximum amplitude value, group average amplitude value, intra-group frequency signal average power value, or total intra-group power value of frequency signals grouped for each frequency band depending on the acoustic signal encoding method. It is also possible to determine whether or not there is a correlation from the ratio between the blocks, and to generate acoustic characteristic comparison information.

  Next, the configuration of another embodiment of the acoustic characteristic comparison unit 12 will be described. FIG. 6 is a block diagram of another embodiment of the acoustic characteristic comparison unit 12 and shows the configuration of the acoustic characteristic comparison unit 12b that acquires acoustic characteristic comparison information according to a predetermined comparison target. The frequency-domain acoustic signal (frequency signal) output from the time-frequency converter 11 in FIG. 2 is supplied to the previous block acoustic signal comparison data buffer 125 in the acoustic characteristic comparator 12b in FIG. The data is temporarily stored for use and supplied to the comparison data generator 126 to generate comparison data.

  Here, the energy change rate for each frequency band of the acoustic signal will be described as a specific example. First, the comparison data generator 126 calculates an energy amount for each frequency band of the frequency signal. By comparing this frequency band with the acoustic signal encoding method used, more accurate characteristics of the acoustic signal can be compared. The energy amount for each frequency band obtained by the comparison data generator 126 is supplied as comparison data to the ratio measuring device 127, where the ratio (change rate) with the energy amount of the previous block from the previous block acoustic signal comparison data buffer 125. ) Is measured by frequency band.

  The ratio determiner 128 determines that the acoustic signals of the current block and the previous block have a high correlation if the average value of the ratio (change rate) measured by the ratio measuring device 127 falls within a predetermined threshold. To do. A determination result based on the ratio determination of the ratio determiner 128 is supplied to the quantization mode selection unit 13 in FIG. 2 as acoustic characteristic comparison information. When the acoustic characteristic comparison information determined to have a high correlation by the ratio determiner 128 is output, the quantization mode selection unit 13 is designated by the acoustic characteristic comparison information to use the low-load quantizer 153 of FIG. Then, the processing of the frequency band weighting information calculation unit 14 is omitted.

  As signals that can be used for the ratio measurement as described above, in addition to the energy amount for each frequency band, the amount of energy in the time domain, the maximum amplitude value, the total energy amount in the frequency domain, the maximum amplitude value of the frequency sample, For example, the maximum amplitude value for each band.

  Next, a procedure for adaptively switching the quantization mode according to the acoustic characteristic comparison information will be described. FIG. 7 is a flowchart showing a procedure in which the acoustic signal encoding apparatus according to the present invention operates adaptively according to the characteristics of the external control signal and the encoded acoustic signal. First, the input time domain digital acoustic signal is blocked at a certain time interval in the time-frequency converter 11 of FIG. 2, and then the digital acoustic signal in the block is divided into a plurality of frequency bands. The signal is converted into a frequency signal for each frequency band (step S1).

  Subsequently, after the acoustic signal characteristic information of the current block (current frame) is generated as described above in the acoustic characteristic comparison unit 12 based on the converted frequency signal or time domain acoustic signal (step S2), the previous The block acoustic signal characteristic information is compared with the current block acoustic signal characteristic information (step S3). Subsequently, the quantization mode selection unit 13 determines whether or not the external control signal (encoding control information in FIG. 2) is a signal designating high-speed encoding processing at the time of the current frame processing (step S4). When it is determined that the signal is for designating the quantization process, it is determined whether or not the low-load quantization of the frequency signal of the current block is possible (step S5). The determination as to whether or not this low-load quantization is possible indicates that the comparison result of the acoustic signal characteristic information of the previous block and the current block obtained in step S3 is an approximate comparison result equal to or less than a preset threshold value. If it is, the quantization accuracy information of the previous block can be applied to the current block, so that it is determined that low-load quantization is possible.

  If low-load quantization is possible, based on the control of the quantization mode selection unit 13, the quantization unit 15 acquires the accumulated previous block quantization accuracy information without performing the band weighting information calculation process. (Step S6) Based on the previous block quantization accuracy information, the frequency signal is immediately quantized by the low-load quantizer 153 (Step S7).

  On the other hand, if it is determined in step S5 that low-load quantization is impossible in spite of the fact that high-speed encoding processing is specified by the external control signal (encoding control information in FIG. 2), the current time zone Without encoding the acoustic signal, the frequency signal and acoustic signal characteristic information converted into the frequency domain of the current block are accumulated (step S8), and the encoding process itself is skipped. At this time, encoding (quantization) of the current block is not performed, and a delay of one block occurs. Therefore, 1 is added to the counter value of the skip counter (initial value 0) as delay information (step S9).

  Also, when it is determined that the external control signal (encoding control information in FIG. 2) is a signal that does not designate the high-speed encoding process during the current frame process (step S4), the normal control process is performed. Is selected so that the high-load quantizer 151 operates, and the band weighting information calculation unit 14 calculates the band weighting information as usual (step S10), and the frequency signal is used for the frequency signal. Then, the high load quantizer 151 executes high load quantization (step S11). The used quantization accuracy information is stored in the quantization information storage 152 in preparation for use of the low-load quantizer of the next block (step S12).

  Although the above operation procedure is a case where no delay occurs, an operation procedure for dealing with a situation in which the encoding process is skipped will be described. FIG. 8 is a flowchart showing an operation procedure that enables the quantization mode to be switched and a plurality of blocks to be processed continuously in response to a situation in which a block delay occurs.

  First, it is determined whether or not the value of the skip counter is greater than 0, that is, whether a block processing delay has occurred (step S21). If the value of the skip counter is 0, there is no block processing delay, so normal encoding processing is performed according to the operation procedure shown in FIG. 7 (no delay) described above (step S22). If it is greater than 0 and a block processing delay has occurred, it is necessary to process a plurality of blocks successively in order to recover the delay.

  Therefore, when the value of the skip counter is 1, for example, first, in the acoustic characteristic comparison unit 12, two adjacent blocks before the current block (here, the value of the skip counter is set) according to the value of the skip counter at that time. 1, the acoustic signal characteristics of the previous block and the previous block are compared (step S23), and based on the comparison result, the quantization mode selection unit 13 is low with the same criteria as in step S5. It is determined whether load quantization is possible (step S24).

  If it is determined that low-load quantization is possible, the low-load quantizer 153 is selected to operate, and the quantization of the two previous blocks when the quantization information storage unit 152 performs the previous quantization is performed. Accuracy information is acquired (step S25), and the frequency signal of the previous block accumulated in step S8 is subjected to low load quantization by the low load quantizer 153 based on the quantization accuracy information (step S26).

  Subsequently, the quantization mode selection unit 13 determines whether or not high-speed encoding processing is designated by an external control signal (encoding control information in FIG. 2) (step S27), and when high-speed encoding processing is designated. The encoding process is terminated because there is no time margin, but if the high-speed encoding process is not designated, it is considered that sufficient time is still left by executing the low-load quantization in step S26. It is done. Therefore, the skip counter is decremented by 1 (step S28), and the process returns to step S21 to perform continuous encoding processing in the subsequent steps. Here, since the value of the skip counter becomes 0 in step S28, the process proceeds from step S21 to S22, and the previous block is encoded according to the flowchart of FIG.

  On the other hand, when the quantization mode selection unit 13 determines in step S24 that low load quantization is not possible, the quantization mode selection unit 13 determines whether high-speed encoding processing is designated by an external control signal (encoding control information in FIG. 2) ( Step S29) When the high-speed encoding process is designated, the frequency signal and acoustic signal characteristic information of the current block that has already been obtained are accumulated without performing the encoding process (Step S30), and the skip counter is incremented by one. (Step S31), the subsequent operation is skipped.

  Further, when it is determined in step S29 that the normal encoding process has been designated by the external control signal (encoding control information in FIG. 2), the quantization mode selection unit 13 causes the high load quantizer 151 to operate. The band weighting information is selected by the frequency band weighting information calculation unit 14 as usual (step S32), and high load quantization is performed on the frequency signal by the high load quantizer 151 (step S32). S33). The used quantization accuracy information is stored in the quantization information accumulator 152 in preparation for use of the low-load quantizer of the next block (step S34).

  As described above, when the value of the skip counter is 1 or more and the high-speed encoding process is not designated, the encoding process is continuously performed for a plurality of blocks in order to recover the block processing delay. Here, assuming that the load applied to the encoding process is 10, the load of the time-frequency converter 11 is approximately 1, that of the band weighting information calculator 14 is 2, and that of the quantizer 15 and other processes is 7. Occupy a proportion. When low-load quantization is performed, the load amount of the bandwidth weighting information calculation unit 14 is 0, and the quantization unit 15 and other processes can be performed with a load of about 1. It will be about 20% more than usual.

  As described above, according to these operation procedures, high-quality sound signal coding is performed at high sound quality while maintaining the integrity of the entire system corresponding to the characteristics of the external control signal (encoding control information) and the sound signal to be encoded. Can be realized.

  The present invention is not limited to the above-described embodiment. For example, the input signal in FIG. 6 may be a time-domain digital acoustic signal that is an input of the time-frequency converter 11. Further, the quantization mode selection unit 13 can also select a quantization mode based on only one of the acoustic characteristic comparison information and the encoding control information from the acoustic characteristic comparison unit 12.

  Furthermore, the present invention includes a computer program that causes a computer to realize the functions of the above-described acoustic signal encoding apparatus. This program may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

It is a block diagram which shows an example of a general acoustic signal encoding apparatus. It is a block diagram of one embodiment of an acoustic signal encoding device according to the present invention. It is a figure which shows an example of the time-amplitude characteristic (for 2048 samples) of an acoustic signal (symphony). FIG. 4 is a frequency characteristic diagram in which 1024 samples before and after the acoustic signal of FIG. 3 are superimposed. It is a block diagram of 1st Embodiment of the acoustic characteristic comparison part in FIG. It is a block diagram of 2nd Embodiment of the acoustic characteristic comparison part in FIG. It is a flowchart explaining an example of operation | movement of this invention apparatus. It is a flowchart explaining the other example of operation | movement of this invention apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 11 Time-frequency converter 2, 14 Frequency band weighting information calculation part 3, 15 Quantization part 4, 16 Encoded signal generation part 12, 12a, 12b Acoustic characteristic comparison part 13 Quantization mode selection part 121 Pre-block sound Signal buffer 122 Phase shift modifier 123 Correlation measuring device 124 Correlation determining device 125 Previous block acoustic signal comparison data buffer 126 Comparison data generator 127 Ratio measuring device 128 Ratio determining device 151 High load quantizer 152 Quantization information accumulator 153 Low Load quantizer

Claims (3)

After the digitized acoustic signal is blocked at regular time intervals, the time domain acoustic signal in the block is divided into a plurality of frequency bands, and the quantum corresponding to the weighting information on the perception for each divided frequency band. In an acoustic signal encoding device that generates an encoded signal by quantizing an acoustic signal in each frequency band with a conversion accuracy,
Compare the acoustic signal characteristics of the previous block and the current block of the digitized acoustic signal in the time domain or the acoustic signal of each frequency band, determine the correlation between adjacent blocks, and provide acoustic characteristic comparison information Obtaining acoustic characteristic comparison means;
First quantizing means for quantizing the acoustic signal of each frequency band of the current block with quantization accuracy according to auditory weighting information for each frequency band;
Second quantization that performs simple quantization of the acoustic signal of each frequency band of the current block using quantization accuracy information when the acoustic signal of each frequency band of the previous block is quantized Means ,
When quantization by the second quantization means is instructed by the encoding control information designating the encoding speed input from the outside , or when the number of skips of the encoding process is 1 or more, the current block Quantization determination means for determining whether or not the acoustic signal of each frequency band can be quantized by the second quantization means based on the degree of correlation indicated by the acoustic characteristic comparison information;
When it is determined by the quantization determination means that the quantization by the second quantization means is impossible based on the acoustic characteristic comparison information, neither the first quantization means nor the second quantization means is selected as the current block. Non-selection means for skipping the encoding process of
When it is determined by the quantization determining means that the second quantizing means can quantize based on the acoustic characteristic comparison information, the second quantizing means is selected and the block before the current block is selected. Quantization selection execution means for performing quantization of the acoustic signal of each frequency band using the quantization accuracy information of the previous block;
The non-selection means includes
When it is determined by the quantization determination means that the quantization by the second quantization means is impossible based on the acoustic characteristic comparison information, the current block used for the quantization execution by the quantization selection execution means Storage means for storing the acoustic signal of each frequency band and the quantization accuracy information according to auditory weighting information for each frequency band;
When it is determined by the quantization determination means that the quantization by the second quantization means is impossible based on the acoustic characteristic comparison information, a counting means for counting the number of skips of the encoding process;
When it is detected that the coding control information indicates the quantization by the second quantization means after the quantization selection by the quantization selection execution means, the value of the counting means is subtracted by a predetermined value. For the quantization by the second quantization means for the acoustic signal of each frequency band of the next block, determination by the quantization determination means is performed, and the coding control is performed after the quantization selection by the quantization selection execution means An encoding control unit for ending the encoding process when it is detected that the information indicates the quantization by the first quantization unit;
Acoustic signal encoding apparatus comprising: a.   The acoustic characteristic comparison means includes a correlation measurement means for obtaining a correlation value between the digitized acoustic signal in the time domain of the adjacent previous block and the digitized acoustic signal in the time block of the current block, and the correlation measurement Correlation determining means for determining whether the correlation value measured by the means satisfies a set threshold and generating the acoustic characteristic comparison information indicating the degree of correlation between the previous block and the current block. The acoustic signal encoding apparatus according to claim 1, wherein:   The acoustic characteristic comparison means is configured such that the total energy in the block of the time domain digitized acoustic signal or each frequency obtained by dividing the time domain digitized acoustic signal into a plurality of frequency bands. Ratio measuring means for measuring the energy ratio between blocks by comparing the total energy in the block of the acoustic signal for each band between the previous block and the current block, and a value corresponding to the energy ratio between the blocks are preset. The acoustic signal encoding apparatus according to claim 1, further comprising: a ratio determination unit that compares the threshold value and generates the acoustic characteristic comparison information indicating a degree of correlation between the previous block and the current block.

Images