KR0134350B1 - Coding and decoding system quantization bit - Google Patents

Abstract

The hatching and decoding system using the adaptive bit allocation of the present invention analyzes the frequency spectrum of an input signal and a buffer for outputting data at a constant rate, and calculates a predetermined threshold level at which humans cannot detect noise due to their auditory characteristics. Threshold level calculating means for receiving information on the data storage amount of the buffer and determining a divided band of the highest frequency to be bit-assigned for quantization, and for the data storage amount of the buffer in the bit-assignable divided bands determined by the threshold level calculating means. An encoding device comprising an adaptive bit allocation according to the adaptive bit allocation, a quantization unit for quantizing the input data according to the adaptively allocated bits, and outputting the input data to the buffer, and an apparatus for receiving and decoding the bit allocation information and the quantized data. By eliminating artificial noises generated in the encoding, the original signal It has the effect of playing a signal close to.

Description

Coding and Decoding System Using Adaptive Bit Allocation

1 is a conceptual diagram illustrating a conventional audio data bit allocation method.

2 is a block diagram of an audio signal encoding apparatus according to an embodiment of the present invention.

3 is a frequency spectrum related to an acoustic signal.

4 is a conceptual diagram for distinguishing three bit allocation schemes according to buffer states.

5 is a conceptual diagram for explaining a method of determining the most divided bandwidth and bit allocation according to a buffer state.

6 is sound data encoded at a variable coding rate.

7 is a block diagram of an acoustic signal decoding apparatus according to an embodiment of the present invention.

The present invention relates to an encoding and decoding system using adaptive bit allocation, and more particularly, compresses data by adaptively converting the number of data bits of an input signal allocated for encoding, and extends the compressed data using bit allocation information. The present invention relates to an encoding and decoding system.

In general, the human hearing range is 20 to 20 kHz. Conventional digital audio products such as CD and DAT use audio characteristics to sample and record and reproduce audio signals with 16-bit resolution at sampling frequencies of 40kHz or higher. In this case, the audio signal can be reproduced at a dynamic range of 96dB and wideband without distortion. When recording and reproducing, transmitting, or receiving a digital audio signal sampled in the above-described manner, more than 700 kbit / sec per channel is required, which is difficult to implement due to the excessive amount of required data. In order to solve this problem, various methods for performing four or more data compressions while maintaining the sound quality of the CD have been proposed.

FIG. 1 is a conceptual diagram showing a conventional audio data bit allocation method, and illustrates a bit allocation method proposed by a moving picture expert group (MPEG).

In general, a phenomenon in which a sound reduces the ability to hear another sound is called a masking effect. The masking effect is used to find the ratio of the level of noise and masking threshold, that is, the level of noise that a human cannot detect even if the noise is mixed with the sound to be heard. When the masking threshold and mask to noise ratio (MNR) for each subband in the minimum time interval (hereinafter, referred to as 'frame') where bit allocation is performed are obtained, the partition having the lowest MNR In the band, 'A' of FIG. 1A, one bit is additionally allocated. In digital signal processing, it is known that the signal to noise ratio (SNR) increases by about 6 decibels (dB) every time the number of bits for representing a sample increases by one bit. Accordingly, the division band 'A' of FIG. 1 (a) to which 1 bit is additionally allocated increases the SNR by about 6 decibels, and the MNR also becomes the decibel value of the dotted line with a 6 dB increase. One bit is allocated to the divided band 'B' having the lowest MNR among the divided bands of FIG. 1 (B) having the changed MNR to increase the MNR. The result of continuing bit allocation until all the bits given in one frame are exhausted is shown in FIG.

When the bit allocation for a given frame is completed by the conventional MPEG method, the MNR is not more than 0 decibels over the entire band even though all the number of bits given by the characteristics of the tones are consumed, and some divided bands are less than 0 decibels. Has a case. In particular, when the MNR of the divided band close to the low frequency range does not reach 0 decibel or more, the listener hears annoying sound when such a tone is reproduced.

An object of the present invention for solving the above problems is to adaptively limit the partition band of the highest frequency among the divided bands for which bit allocation is considered according to the buffer state, and to limit the bit allocation of the divided bands for which bit allocation is considered. The present invention provides an encoding and decoding method capable of reproducing a sound close to an original sound during reproduction by varying the buffer state.

Another object of the present invention for solving the above problems is to provide an encoding and decoding apparatus implementing the above method.

An object of the present invention as described above is to divide a frequency band of an input signal into a plurality of divided bands, and to spectrally analyze the input signal to calculate a predetermined threshold level at which noise cannot be reduced due to human sensory characteristics. In the encoding method for quantizing the signals of the respective divided bands by using, storing the quantized and multiplexed data in a buffer, generating information about the amount of data storage of the buffer, Generating information on the amount of data storage, and determining a final division band, which is a division band of the highest frequency, among the allocation bands according to the temporal change of the data storage amount, and assigning a bit to the division bands larger than the final division band. A threshold level readjustment step of adjusting a threshold level so as not to cause error; and storing the readjusted threshold level and buffer data. Adaptively allocating the number of bits required for quantization to a divided band having a frequency less than or equal to the last divided band according to a quantity; and quantizing a signal of each divided frequency band according to the allocated number of bits. Receiving a signal encoded by the above-described encoding method, separating the encoded data bit allocation information by demultiplexing the encoded signal, and separating the separated bits. Dequantizing the encoded data using allocation information, and synthesizing the data of the dequantized frequency division bands.

Another object of the present invention is to provide an encoding apparatus comprising means for dividing a frequency band of an input signal into a plurality of divided bands, and means for analyzing a frequency spectrum of the input signal. A buffer for generating information on the storage amount and a predetermined threshold level at which noise cannot be detected due to human sensory characteristics according to the spectrum of the input signal obtained from the spectrum analysis means, and according to the temporal change of the data storage amount of the buffer. A threshold level calculating means for determining a final division band, which is a division band of the highest frequency among the bit-assigned division bands, for each frame, which is a unit of data processing, and a predetermined threshold level of the occupancy ratio and the threshold level calculation means of the buffer; A first bit adaptively allocating the number of bits required for quantization to divided bands below the last divided band And an quantization means for quantizing a signal of each divided band output from the frequency band dividing means and outputting the signal to the buffer according to the bit allocation information provided by the bit allocation means. A coding apparatus using the sequential encoding apparatus, an input terminal for receiving a signal encoded by the above-described coding apparatus, a means for demultiplexing a signal applied to the input terminal with decoded data, and the separated bit allocation information. Means for providing to the dequantization means, and means for synthesizing a signal of each dequantized frequency component.

Hereinafter, a preferred embodiment of implementing an encoding and decoding system using adaptive bit allocation according to the present invention will be described in detail with reference to the accompanying drawings.

2 is a block diagram of an audio signal encoding apparatus according to an embodiment of the present invention, and illustrates an audio signal encoding apparatus that receives an analog sound signal and outputs the sound data at a constant transmission bit rate.

The apparatus of FIG. 2 includes an A / D converter 21 for converting an analog signal into digital data, a fast Fourier transform (FFT) circuit, and a conversion for separating the frequency components of the input data for each divided band. A band splitter 22, a spectrum analyzer 24 for analyzing the frequency spectrum of the input data on a frame-by-frame basis, a first buffer 26 for outputting data input at a variable data rate at a constant data rate, and A threshold level calculator 25 which receives the buffer occupancy and the output data of the spectrum analyzer 24, which is information on the data storage state of the first buffer 26, and calculates the MNR considering the buffer state; The first bit allocator 27 which receives the buffer occupancy rate and the output data of the threshold level calculator 25 and allocates the bit data to the respective divided bands, and the output data of the conversion and band divider 22 bit. Quantization unit 23 to quantize according to the allocation information Combining the output data and the bit allocation information of the quantization unit 23 will be provided with a format conversion unit 28 for output to the first buffer (26).

When the sound signal is input to the A / D converter 21 through the input terminal a, the A / D converter 21 receives a sampling frequency f _s of a predetermined size and samples the above sound signal into digital data. Convert the output. The conversion and band dividing unit 22 performs fast Fourier transform of the input data into data in the frequency domain, and then frequency-band divides the acoustic data in each frame and outputs the frequency data to the quantization unit 23. The spectrum analyzer 24 receives the acoustic data represented as a function of time, performs fast Fourier transform, analyzes the frequency spectrum of the acoustic data in each frame, and outputs it to the threshold level calculator 25.

FIG. 3 shows frequency spectrums related to sound signals, and shows frequency spectra of sound signals belonging to one frame in divided band units divided by dotted lines. The sound pressure level (SPS) and the masking threshold of the sound signal are indicated by the solid line and the dashed line in FIG. In FIG. 3B, the signal-to-mask ratio (SMR) is indicated by a dotted line, and the MNR is indicated by a solid line.

In FIG. 3, SMR is obtained by subtracting the masking threshold in SPL, and MNR is obtained by subtracting SMR from SNR. When the bit allocation is not made, the SPL and the noise level coincide with each other, and MNR = SMR from MNR = SNR-SMR. Thus, with no bit assignment at all, the MNR and SMR have decibels (dB) that are symmetrical to each other on the frequency axis f, which is indicated in the frequency domain above 'A' in FIG. .

Since the human ear is more sensitive to low frequency components than high frequency components, in the present invention, when the bit allocation is completed, the division band having the MNR of 0 decibel or less does not appear in the low frequency region and the middle frequency region. Do not allocate bits. Hereinafter, the highest frequency division band for which bit allocation is performed is referred to as a 'final division band', and the final division band determination of the threshold level calculating unit 25 and the calculation of the MNR and the variable bits of the first bit allocation unit 27 are performed. The allocation method will be described with reference to FIGS. 4 and 5 below.

4 is a conceptual diagram for distinguishing three bit allocation methods according to a buffer state. The left end is defined as an empty state, and the right end is defined as a state where data is completely filled in the buffer. A, B, and C are used to divide the data storage state or buffer occupancy of the buffer into three areas. The unit interval divided by the dotted lines in the area B is the number of data bits in one frame determined by the constant output transfer rate of the first buffer 26.

FIG. 5 is a conceptual diagram illustrating final division band determination and bit allocation according to the buffer state. FIG. 5 is a MNR when final division band determination is completed, and FIG. 5 is a diagram in which bit allocation is completed. The final final MNRs are shown respectively.

The threshold level calculator 25 receives the information on the data storage state of the first buffer 26 by the acoustic data of the previous frame and the frequency spectrum of the current frame output from the spectrum analyzer 24. The considered MNR is calculated as follows. First, the threshold level calculator 25 calculates the MNR indicated by the solid line in FIG. 5A by using the acoustic data frequency spectrum output from the spectrum analyzer 24. Then, the data of the previous frame is passed through the first buffer 26 to receive the buffer state information generated to adaptively change the last divided band of the current frame to be considered bit allocation. That is, when the amount of data stored in the first buffer corresponds to the region B of FIG. 4, the data storage amount of the first buffer 26 in the previous frame is compared at a comparison point of a plurality of data amounts divided by the data interval of one frame. . By comparison, if the data storage amount of the first buffer 26 is lower than the predetermined comparison reference point and exceeds the reference point, the comparison is performed to compare the data storage amount in the next frame by moving the final division band to the low frequency side by one division band. The reference point is moved to a reference point with a large data storage by two intervals. On the other hand, if the data storage amount is higher than the predetermined reference point by the inspection and falls below the reference point, the final division band is moved toward the high frequency side by one division band, and the comparison reference point is moved to the lower side of the data storage amount by one price. The SPL is set to a very low value in the division bands higher than the final division band so as not to assign the bit allocation to the division bands whose frequency is larger than the final division band in which bit allocation is considered. In the divided bands where the SPL is set to a very low value, a new MNR is formed by the above-described relation MNR = -SMR, and the SPL of the divided bands higher than the last divided band is set to -30 days, Ignoring the masking addition effect due to the tonal and nontonal components present in the divided bands, bit allocation is applied to the divided bands in the frequency domain above A by replacing the MNR indicated by the dotted line in FIG. Not done. Here, the reason for ignoring the masking addition effect due to sound quality and noise components present in the division band of A or more is to not affect the division band of less than A.

In addition, the threshold level calculating unit 25 initializes the final divided band every 50 frames (variable) in order to prevent an error in the determination of the super-divide band due to the sudden change in the data amount, and the initializing value of the MNR is 0 decibel or less. The split band of the highest frequency is determined as the final split band. In addition, when the signal changes drastically, the lowest frequency division band that the final division band can have is set, and when the obtained final division band is lower than the predetermined frequency division band, the final frequency allocation where the bit allocation is performed is performed. Determined as the split band. If the difference between the last divided band newly determined every 50 frames and the last divided band of the immediately preceding frame is three or more divided bands, the divided two bands are moved in the high frequency direction so as to be closer to the last divided band of the immediately preceding frame. The band is determined as the final divided band initialized every 50 frames.

When the threshold level calculator 25 calculates and outputs the MNR by the above-described method, the first bit assigner 27 bits the frequency division bands below the final division band determined by the threshold level calculator 25. Perform the assignment. The bit allocation method of the first bit allocation unit 27 is divided into three methods corresponding to A, B, and C according to the data storage state of the first buffer 26. When the first buffer 26 has a data storage amount or occupancy corresponding to area A in FIG. 4, the first bit allocation unit 27 underflows in the first buffer 26 that outputs data at a constant transfer rate. The minimum number of bits per frame of the sound data encoded and input to the first buffer 26 is equal to the transmission oil of the output data of the first buffer 26 so as not to occur. The maximum number of bits per frame of the acoustic data is not more than 1.5 times the data rate of the first buffer 26. When the data storage amount of the first buffer 26 belongs to the area B of FIG. 4, the minimum number of bits per frame of tone data encoded and input to the buffer 26 does not define a limit range, and the maximum number of bits is set to the first range. The data rate of the buffer 26 is limited to 1.5 times. When the storage amount of the buffer 26 belongs to the region C of FIG. 4, the minimum number of bits per frame of data input to the first buffer 26 does not define a limit range, and the maximum number of bits is the data of the buffer 26. The transfer rate is controlled so that no overflow occurs in the buffer 26.

The first bit allocation unit 27 selects one of the bit allocation methods described above according to the data storage state of the first buffer 26, and combines the MPEG method of FIG. The process of bit allocation to the divided band having the minimum decibel value among the divided bands equal to or less than 0 decibel is repeated. If the MNR of all the divided bands is 0 decibel or more in the bit allocation process for frequency band below 'A' in Fig. 5 (a), the bit allocation is terminated. Determine the number of encoded bits. Accordingly, the first bit allocation unit 27 allocates all the divided bands having an MNR of 0 decibel or less as shown in FIG.

The quantization unit 23 receives bit allocation information of the first bit allocation unit 27 and quantizes sound data output from the conversion and band division unit 22. When the quantized sound data is input to the formatter 28, the formatter 28 multiplexes the bit allocation information and the quantized sound data transmitted from the bit assigner 27 to the first buffer 26. Will output

6 shows sound data encoded at a variable coding rate, which shows that the data length of each frame is variable.

Since the acoustic data stored in the first buffer 26 is input with variable lengths for each frame as shown in FIG. 6, the first buffer 26 remains untransmitted due to the constant data rate. The data is stored, and information on the data storage state is transmitted to the threshold level calculating section 25 and the first bit allocation section 27 so that bit allocation can be performed during the quantization process of sound data of the next frame.

FIG. 7 is a block diagram of a fornication signal decoding apparatus according to an embodiment of the present invention, and shows an apparatus for restoring sound data encoded and transmitted by the apparatus of FIG.

The apparatus of FIG. 7 is connected to the data output terminal of the second buffer 71 and receives an inverse quantization unit for dequantizing the sound data represented by the frequency component from the deformatter 72 which demultiplexes the input data. 74) and a second bit allocator 73 for receiving bit allocation information output from the inverse formatter 72 described above and supplying bit allocation information for nutritionalization of acoustic data; D / A converter for outputting analog data using the conversion and split band synthesis unit 75 and the sampling frequency f _s for converting and outputting each split band into sound data expressed as a function of time. 76).

When sound data transmitted from the encoding apparatus of FIG. 2 at a constant data rate is stored in the second buffer 71, the second buffer 71 outputs data in units of frames to the deformatter 72. FIG. The inverse formatter 72 demultiplexes input data, that is, additional information including bit allocation information, and coded sound data. The second bit allocator 73 receives the bit allocation information output from the inverse formatter 72, and the bit quantization information of each frame is input by the inverse quantizer 74 to receive the input sound data and the corresponding bit allocation information. When the inverse quantized data is output, the conversion and division band synthesis unit 75 synthesizes each frequency division band of the input sound data and inversely performs fast Fourier inverse conversion on the synthesized sound data of the frequency components. The acoustic data represented as a function of time by the split band synthesis and the inverse transform is output by the D / A converter 76 as an analog sound signal. Therefore, the sound signals encoded and decoded by the apparatus of FIGS. 2 and 7 are reproduced as signals close to the original sound.

The numerical values set in the above embodiment do not limit the technical content of the present invention, and limit the final divided band according to the buffer state of the encoder and a plurality of different values according to the data storage amount of the buffer of the encoding side in the allocated bits. Any type of device for performing bit allocation in other ways may be implemented in the technical scope of the present invention. In the above embodiment, the present invention has been described using encoding of an audio signal. However, the encoding and decoding system of an image signal using a human visual characteristic insensitive to a high frequency region of a luminance signal for determining an image quality felt by a human is provided. Implementation is also possible within the technical scope of the present invention.

According to the encoding and decoding system using the adaptive bit allocation of the present invention as described above, the partition band of the highest frequency allocated to the bit is determined by using the data storage state of the buffer for storing the encoded data, and the data storage state of the buffer. By adaptively allocating bits to data to be encoded according to the present invention, there is an effect of removing artificial artifacts generated in encoding and reproducing a signal close to the original signal before encoding.

Claims (19)

The frequency band of the input signal is divided into a plurality of divided bands, and spectrum analysis of the input signal yields a predetermined threshold level at which noise cannot be detected due to human sensory characteristics, and the signal of each divided band is used using the threshold level. A coding method for quantizing a signal, the method comprising: storing quantized and multiplexed data in a buffer, generating information about a data storage amount of the buffer, and performing a bit allocation according to a temporal change of the data storage amount A threshold level readjustment step of determining a final division band, which is a division band of the highest frequency of the bands, and adjusting a threshold level so as not to allocate bits to the division bands larger than the final division band, and the readjusted threshold level and buffer data. The number of bits required for quantization is allocated to a divided band having a frequency less than or equal to the last divided band according to the storage amount. Assigning a eungjeok and coding method comprising the steps of: quantizing a signal of each frequency band, which is divided according to the number of the allocated bits. 2. The method of claim 1, wherein the threshold level re-adjustment comprises dividing the data into a variable region of the last divided band and a region of a predetermined final divided band according to a data storage amount of a buffer. 3. The method of claim 2, wherein the critical level readjustment step moves the final division band to one low frequency side by one division band when the data storage amount of the buffer increases by more than a predetermined size in the divided band variable region, and the data storage amount of the buffer is greater than or equal to another predetermined size. And if it is decreased, the final division band is shifted toward the lower frequency by one division band. 4. The encoding method according to claim 3, wherein the threshold level readjustment step resets the last divided band of the last divided band variable region at predetermined intervals. 3. The method of claim 2, wherein the adaptive bit allocation step uses different bit allocation ranges in a region in which the final divided band is variable and a region in which the final divided band is constant. A method of decoding an encoded signal, the method comprising: receiving a signal encoded by the method of claim 1, demultiplexing the encoded signal to separate encoded data and bit allocation information, and separating the encoded signal Dequantizing the encoded data using the allocated bit allocation information, and synthesizing data of the respective dequantized frequency division bands. The decoding method according to claim 6, wherein the inverse quantization step uses the number of bits allocated to predetermined data in the quantization step of encoding to dequantize the same data. An encoding apparatus comprising means for dividing a frequency band of an input signal into a plurality of divided bands, and means for analyzing a frequency spectrum of the input signal, comprising: a buffer for receiving encoded data and generating information on the amount of data storage; According to the spectrum of the input signal obtained by the spectrum analysis means calculates a predetermined threshold level that can not detect noise due to human sensory characteristics, and the highest frequency of the divided bands allocated bit according to the temporal change of the data storage amount of the buffer Exception level calculation means for determining a final division band, i.e., a division band of?, For each frame, which is a unit of data processing, and quantizing the divided divisions below the last division band by receiving a share of the buffer and a predetermined threshold level of the threshold level calculation means. First bit allocation means for adaptively allocating the number of bits necessary for And quantization means for quantizing a signal of each divided band output by the frequency band dividing means and outputting the signal to the buffer according to the bit allocation information provided by the stage. 10. The apparatus of claim 8, wherein the threshold level calculating means divides the final divided band into a variable region and a constant region according to a data storage amount of a buffer. 10. The data storage apparatus of claim 9, wherein the threshold level calculating means moves the final divided bandwidth by one division band toward the lower frequency when the data storage amount of the inter-frame buffer increases by more than a predetermined size in a region where the final divided bandwidth is variable. And an additional bit is reduced by more than the other predetermined size. 12. The method of claim 10, wherein the comparison criteria for increasing the data storage amount of the buffer is set at twice the data transfer rate of the buffer, and the comparison criteria for reducing the data storage amount of the buffer is set at the data transfer rate of the buffer. Encoder using bit allocation. 12. The encoding device according to claim 10, wherein the threshold level calculating means resets the last divided band of the final divided band variable region at predetermined time intervals. The coding method according to claim 12, wherein the final divided band reset at each predetermined time interval is a split band of the highest frequency at which the mask-to-noise ratio (MNR), which is the threshold level, is 0 decibel or less. Device. 13. The method of claim 12, wherein a split band of the lowest frequency that the final split band has in advance is set in advance, and the split band of the set lowest frequency is set if the final split band that is reset at each predetermined time interval is less than or equal to the preset band. A coding apparatus using adaptive bit allocation, characterized in that the final division band is determined. 15. The method of claim 14, wherein if the divided band interval between the last divided band to be reset at a predetermined time interval and the last divided band of the immediately preceding frame is three or more divided bands, the last divided band to be reset in the last divided band of the immediately preceding frame. And a split band shifted by two split bands toward the final split band. The encoding device using adaptive bit allocation according to claim 9 or 10, wherein the first bit allocation means determines differently the bit allocation ranges of the region where the final division band is variable and the constant region. 17. The predetermined number of bits according to claim 16, wherein the bit allocation range in the region in which the final divided band is variable is a predetermined number of bits in which the maximum number of bits per frame of data encoded and input to the buffer is greater than the number of bits per frame by the data rate of the buffer. And an encoding apparatus using adaptive bit allocation. 17. The method of claim 16, wherein the final divided band is divided into a first region close to a state in which the buffer is completely empty and a second region close to a state in which the buffer is completely filled. bits are allocated to input data into the buffer to prevent the occurrence of underflow. In the second region, bits are allocated to input data into the buffer to prevent overflow from the buffer. An encoding apparatus using adaptive bit allocation. An apparatus for decoding an encoded signal, comprising: means for separating the encoded data and bit allocation information by demultiplexing an input terminal receiving the signal encoded by the apparatus of claim 8 from a signal applied to the input terminal; And means for inversely generating the separated encoded data, means for providing the separated bit allocation information to the inverse quantization means, and means for synthesizing a signal of each dequantized frequency component. Decryption device.