KR0181061B1 - Adaptive digital audio encoding apparatus and a bit allocation method thereof - Google Patents

Abstract

The novel apparatus capable of adaptively encoding an input digital audio signal comprises: a first meter for measuring signal-to-mask ratio data, sound pressure levels and masking thresholds for respective subbands of the digital audio signal; Measure the amount of cognitive information for each frame of the input digital audio signal based on the measured signal-to-mask ratio data, sound pressure levels and masking thresholds, and at least two or more current lights corresponding to the measured cognitive information amounts. A second meter for measuring mean and standard deviation parameters for a frame group containing previous frames; A bit for each subband is adaptively determined based on the measured signal-to-mask ratio data, cognitive information quantities, and average and standard deviation parameters, and a bit corresponding to the determined bit for each subband. Bit allocation means for generating allocation information; A quantizer for quantizing the filtered subband audio signals in response to bit allocation information for the generated subbands; And circuitry for formatting the quantized audio signal and the generated bit allocation information.

Description

Adaptive Digital Audio Coding Device and Its Bit Allocation Method

1 schematically illustrates a novel apparatus and bit allocation method for adaptively encoding an input digital audio signal according to the present invention.

FIG. 2 is a detailed block diagram of the first bit allocation unit shown in FIG.

* Explanation of symbols for main parts of the drawings

10: subband filtering device

20, 30: first and second cognitive parameter measuring instrument

40, 50: first and second bit allocation units

60: quantum 70: formatting circuit

TECHNICAL FIELD The present invention relates to a digital audio encoding method and apparatus therefor, and more particularly, to an improved adaptive digital encoding of an input digital audio signal having a plurality of frames based on cognitive information amounts of frames corresponding to an auditory characteristic by a human being. An audio encoding device and a bit allocation method thereof.

Transmit digitalized audio signals to achieve high sound quality, such as Compact Disc (CD) and / or Digital Audio Tape (DAT) players. Audio signals are represented in digital form and, in particular, when used in high definition television (HDTV), a significant amount of data is transmitted. However, because the available bandwidth allocated to audio signals is limited, a significant amount of data, such as a 16-bit pulse code modulation (PCM) audio signal sampled at 48 KHz, i.e. 768 Kbits per second, is limited to a limited audio band, e.g. For example, to transmit at about 128KHz, it must be compressed.

Among various audio compression devices or techniques, a so-called Moving Pictures Expert Group (MPEG) -audio algorithm has been proposed that uses Psychoacoustic Algorithm for HDTV applications.

MPEG-audio algorithms consist of four main components: subband sampling, psychoacoustic algorithms, quantization and coding, and frame formatting. Subband filtering is the process of mapping the input PCM audio signal from the time domain to the frequency domain. A filter bank with B (eg 32) subbands may be used. In each subband, 12 or 36 samples are grouped and processed, grouped samples from the B subbands, ie, B x 12 or 36 samples, in one frame. Each frame is a processing unit for encoding, transmitting and decoding an audio signal. Psychoacoustic modeling generates a set of data, eg, signal-to-mask ratio (SMR) data, for each subband or group of subsamples to control the processing of quantization and encoding. Subsequently, in the step of quantizing and encoding the subband samples, the available bits are quantized by adaptively allocating available bits to each subband of one frame with reference to the generated SMR data. The frame format formats and transmits the frame data in an appropriate form along with other necessary additional information.

However, since the aforementioned MPEG audio technique allocates a fixed bit to each frame, the technique includes averages and standard deviations of the input digital audio signal that may be changed continuously between the frames. However, there is a problem in that statistical characteristics such as perceptual entropy cannot be reflected.

Accordingly, an object of the present invention is to provide a novel apparatus and its bit, which adaptively encode an input digital audio signal having a plurality of frames based on cognitive information amounts of frames corresponding to a human auditory characteristic, thereby improving coding efficiency and sound quality. To provide an allocation method.

According to an aspect of the present invention, an apparatus for adaptively encoding an input digital audio signal having a plurality of frames, each frame comprising a plurality of subbands, wherein the input digital audio signal is filtered to filter the input digital audio signal. Means for dividing into band units; First measuring means for measuring signal-to-mask ratio data, sound pressure levels and masking thresholds for respective subbands of the input digital audio signal; Measure cognitive information amounts for each frame of the input digital audio signal based on the measured signal-to-mask ratio data, sound pressure levels and masking thresholds, and at least two currents corresponding to the measured cognitive information amounts. Second measuring means for measuring average and standard deviation parameters for a frame group comprising previous frames; Adaptively determine bits for each of the subbands based on the measured signal-to-mask ratio data, cognitive information quantities, and mean and standard deviation parameters, and correspond to the bits determined for the respective subbands. Bit allocation means for generating bit allocation information; Means for quantizing each filtered subband audio signal in response to bit allocation information for each generated subband; And means for formatting the quantized audio signal and the generated bit allocation information.

According to another aspect of the invention, a bit allocation method is used to adaptively encode an input digital audio signal having a plurality of frames, each frame comprising a plurality of subbands, the method comprising: the input digital; Filtering and dividing an audio signal into a plurality of subband units; Measuring signal-to-mask ratio data, sound pressure levels, and masking thresholds for respective subbands of the digital audio signal; Cognitive information amounts for each frame of the input digital audio signal are measured based on the measured mask ratio data, sound pressure levels and masking thresholds, and at least two current and previous ones corresponding to the measured cognitive information amounts. Measuring average and standard deviation parameters for a frame group comprising the frames; Measuring decision levels of the frame group based on the measured mean and standard deviation parameters; Determine bits for each frame of the frame group based on the measured decision levels, the total number of decision levels, the amount of cognitive information, and a predetermined average bit, and the bits determined for each of the frames. Generating bit allocation information corresponding to the; And determine bits for respective subbands of each frame based on the measured signal to mask ratio data and the generated bit allocation information, and bits corresponding to the bits determined for the respective subbands. Generating allocation information.

The above and other objects and features of the present invention will become apparent from the description of the preferred embodiments provided in connection with the accompanying drawings.

Referring to FIG. 1, there is shown a block diagram schematically illustrating an adaptive digital audio encoding apparatus and a bit allocation method according to the present invention.

The adaptive digital audio encoding apparatus includes a subband filtering device 10, first and second cognitive parameter measuring instruments 20 and 30, first and second bit allocation units 40 and 50, quantizer 60, and formatting. Circuit 70.

The input digital audio signal X (n) of the i-th frame, or current frames, containing N samples (i.e., n = 0,1, .. N-1, where N is a positive integer) Is applied to the first perceptual parameter meter 20 and the subband filtering device 10. The subband filtering device 10 performs a subband filtering operation of the input digital audio signal. As used herein, a frame represents a portion of a digital audio signal corresponding to a fixed number of audio samples, and is also a processing unit for encoding and decoding the digital audio signal.

As shown, the subband filtering device 10 receives the input digital audio signal of the current frame to provide subband filtering techniques well known in the art, for example, ISO / IEC JTCI / SC2 / WG 11, Part 3 By using the method disclosed in the so-called MPEG audio algorithm described in Audio Proposal, CD-11172-3 (1991), filtering of input digital audio signals is performed. That is, the subband filtering device 10 divides an input digital audio signal having a sampling frequency fs into subbands having B (eg, 32) identical frequency bands having a sampling frequency fs / B and Provide the divided subband audio samples to the quantizer 70.

Meanwhile, the first cognitive parameter measurer 20 receives the input digital audio signals of the current frame and, for example, each of the subbands included in the input digital audio signal using the psychoacoustic model discussed in the MPEG audio algorithm. Signal to Mask Ratio (SMR) data, Sound Pressure Levels, and Masking Thresholds are measured for. SMR data for each subband is extracted as follows.

Where j is the subband index (j = 0, 1, .., B-1), B is the total number of subbands in one frame, and SMR (j) is the signal-to-mask ratio in the jth subband. ; P (j) is the sound pressure level in the jth subband measured from the Fast Fourier Transform technique; M (j) _ is the masking threshold in the jth subband; The units of SMR (j), P (j) and M (j) are all decibels (dB).

The masking threshold represents the audible limit, which is the sum of the inherent audible limit or negative threshold and the increment generated by the presence of other pure and non-pure components of the audio signal. Then, SMR1 (j) data is provided to the second bit allocation unit 50, while P (j) and M (j) are supplied to the second cognitive parameter meter 30.

The second cognitive parameter measurer 30 measures the amount of cognitive information PE (i) for the input digital audio signal of the i-th frame based on the sound pressure levels and the masking thresholds from the first cognitive parameter measurer 20. do. The amount of recognition information PE (i) for the input digital audio signal of the i-th frame may be expressed as follows, as is well known in the art.

Where i, j and B have the same meanings as defined above. Equation (2) can be obtained by applying the so-called Rate Distortion Theory, which corresponds to the amount of cognitive information based on human auditory characteristics. The second cognitive parameter meter 30 then measures the amount of cognitive information measured for Q (e.g. 4) current and its previous frames, i.e., PE (i), PE (i-1), PE. By grouping (i-2) and PE (i-3), bits between frames grouped according to the processing of the first bit allocation unit 40 to be described in detail below with reference to FIGS. 1 and 2. And the average (PE) and standard deviation (PEstd) parameters representing their statistical characteristics, using the total amount of cognitive information of the frame group. The average parameter PEm for the total amount of cognitive information of the frame group may be obtained as follows, as is well known in the art.

Where p (p = 0,1, .. Q-1) is the frame index used in the frame group, Q is the total number of frames in the frame group, and PE (p) is the pth in the frame group. Represents the amount of cognitive information of the frame.

Therefore, the standard deviation PEstd for the total cognitive information amounts of the frame group may be calculated as follows, as is well known in the art.

Where p and Q have the same meanings as defined above.

The recognition information amount PE (p) and the average and standard deviation parameters PEm and PEstd of the p-th frame grouped and measured by the second recognition parameter measurer 30 are applied to the first bit allocation unit 40. . The first bit allocation unit 40 determines a bit for each frame included in the frame group based on the amount of acknowledgment information of the pth frame from the second recognition parameter measurer 30 and the average and standard deviation parameters. The bit allocation information FBI corresponding to the bits determined for the respective frames of the frame group is provided to the second bit allocation unit 50 and the formatting circuit 70.

Referring to FIG. 2, a detailed block diagram of the first bit allocation unit 40 shown in FIG. 1 is shown. The first bit allocation unit 40 includes a decision level measurer 41, a subtractor 42, and a bit allocation device 43.

The decision level measurer 41 calculates an average from the second cognitive parameter measurer 30 shown in FIG. 1 so that the bit allocation device 43 adaptively allocates a bit to each frame in the frame group. And the optimal determination levels of the frame group based on the standard deviation parameters PEm and PEstd. In a preferred embodiment of the present invention, the kth determination level D (k) of the frame group can be expressed as follows.

Where k (k = −q ~ q) is the decision level index, q is a positive integer and NF is the normalization factor in the frame group.

As can be seen from equation (5), the level interval between the kth decision level D (k) and the (k-1) th decision level D (k-1) of the frame group is the second parameter. In contrast to the standard deviation parameter PEstd from the meter 30 and the normalization factor NF of the frame group, the total number of decision levels is preset. The total number of decision levels may be determined based on the coding efficiency and sound quality of the encoding apparatus. The normalization factor (NF) of the frame group commonly used in the decision level measurer (41) is obtained from the second cognitive parameter measurer (30) in order to extract the optimal decision levels of the frame group that closely match the actual human auditory characteristics. It can preferably be determined by using the mean and standard deviation parameters, and the global mean PEgm pre-stored in its memory (not shown) and the mean PEgstd of the global standard deviation parameters. Each of the global parameters and the mean parameters of the global standard deviations can be easily measured by using the mean and standard deviation parameters measured during the predetermined period. In a preferred embodiment of the present invention, the normalization factor (NF) of the frame group can be determined as follows.

In Eq. (6), PEm and PEstd are mean and standard deviations of cognitive information obtained from Eq. (3) and Eq. (4), and PEgm and PEgstd are the average values of PEm and PEstd for a long time (for example, 4 sec). Correspondingly, this long-term characteristic of the audio signal is generally constant, so that the previously stored value can be used, or the previous PEm and PEstd can be averaged for a long time.

As can be seen from equations (5) and (6), note that the decision levels of the frame group can be determined as integer multiples of the average parameter.

On the other hand, the subtractor 41 subtracts the recognition parameter E (p) from the second recognition parameter measuring instrument 30 by the average parameter PEm, thereby causing the deviation signal E (p) of the p-th frame in the frame group. Calculate Then, the determination level D (k) and the total number of decision levels (ie 2q + 1) measured at the decision level meter 41 are simultaneously provided to the bit allocation device 43.

The bit allocation device 43 divides the bits for each frame group of the frame group based on the decision levels from the decision level meter 41 and the total number of decision levels, and the deviation signal from the subtractor 42. And the bit allocation information corresponding to the determined bits for each frame are respectively provided to the second bit allocation unit 50 and the formatting circuit 70 shown in FIG. In a preferred embodiment of the present invention, the bit allocation (FB (p)) of the p-th frame in the frame group can be determined as follows.

Where p has the same meaning as defined above, where FBm is the average bit per frame (e.g., 3072 bits (i.e., 48KHz sampling frequency, 16-bit PCM form of audio data, and therefore 128Kbits per second). )), BV represents a predetermined bit shift value, 2q + 1 represents the total number of predetermined decision levels, and I is the level index in the pth frame.

As can be seen from equation (7), the bit allocation FB (p) for the p-th channel can be calculated by adding the average bit FBm and the variable bit calculated from its second term. The predetermined bit shift value BV may be determined to be equal to the average bit value for one frame as defined in Equation (7), and the level index I for the pth frame in the frame group is determined by the determination level meter. Can be determined based on the decision levels D (k) from 41 and the deviation signal E (p) from the subtractor 42. In a preferred embodiment of the present invention, the level index (I) for the pth frame in a frame group may be represented as shown in the following table (where the spacing of decision levels is 1.27, and the decision level index k is -2 to 2).

As can be seen from the table, for example, if the deviation signal E (p) exists between the decision levels (2.55 to -1.28), the level index (I) of the p th frame is -1. And, if the deviation signal is between decision levels (1.27 to -1.26), the level index can be selected as zero. In this way, the bit allocation of the pth frame can be advantageously determined by using equation (7).

Then, in the bit allocation unit 45, the bit allocation information FBI corresponding to the determined bit for each frame of the frame group and the signal mask mask from the first recognition parameter measurer 20 shown in FIG. The ratio SMR (j) is simultaneously provided to the second bit allocation unit 50, and the bit allocation information is provided to the formatting circuit 70.

Referring back to FIG. 1, the second bit allocation unit 50 receives the signal-to-mask ratio supplied from the first recognition parameter measuring unit 20 and the bit allocation information provided from the first bit allocation unit 40. And determining bits for each subband included in each frame of the frame group, and simultaneously providing bit allocation information (SBI) corresponding to the bits determined for each subband to the quantizer 60 and the formatting circuit 70 at the same time. do. The process principle used in the second bit allocation unit 50 is based on the total signal-to-mask ratio for one frame within the range not exceeding the available bits for each frame transmitted from the first bit allocation unit 40. It depends. Then, the bit allocation information SBI for each subband from the second bit allocation unit 50 and the divided subband audio samples from the subband filtering device 10 are transferred to the quantizer 60. Provided at the same time.

The quantizer 60 adaptively quantizes the divided subband audio samples from the subband filtering device 10 based on the bit allocation information corresponding to the audio samples from the second bit allocation unit 50. To the formatting circuit 70.

In the formatting circuit 70, for formatting and transmitting the quantized audio samples from the quantizer 60 and the bit allocation information from the first and second bit allocation units 40 and 50. By transmitting to a transmitter (not shown), the coding efficiency and sound quality of the input digital audio signal can be improved. The principle and function of the subband filtering device 10, the second bit allocation unit 50, the quantizer 60 and the formatting circuits 70 are basically the same as those discussed in the MPEG audio algorithm.

Although the present invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes may be made without departing from the spirit and scope of the appended claims.

Claims (5)

An apparatus for adaptively encoding an input digital audio signal having a plurality of frames, each frame comprising a plurality of subbands, comprising: means for filtering and dividing the input digital audio signal into a plurality of subband units; First measuring means for measuring signal-to-mask ratio data, sound pressure levels and masking thresholds for respective subbands of the input digital audio signal; Measure cognitive information amounts of respective frames of the input digital audio signal based on the measured signal-to-mask ratio data, sound pressure levels and masking thresholds, and at least two current and previous ones corresponding to the measured cognitive information amounts. Second measuring means for measuring average and standard deviation parameters for a frame group comprising frames; Adaptively determine bits for each of the subbands based on the measured signal-to-mask ratio data, cognitive information quantities, and mean and standard deviation parameters, and correspond to the bits determined for the respective subbands. Bit allocation means for generating bit allocation information; Means for quantizing each filtered subband audio signal in response to bit allocation information for each generated subband; And means for formatting the quantized audio signal and the generated bit allocation information. 2. The apparatus of claim 1, wherein the bit allocation means comprises: means for measuring decision levels for the frame group based on the measured mean and standard deviation parameters; Means for generating a deviation signal indicative of a deviation between the measured cognitive information amount and the average parameter; Determine a bit for each frame of the frame group based on the measured decision levels, the total number of decision levels, the amount of cognitive information, and a predetermined average bit, First bit allocation means for generating corresponding bit allocation information; And determining bits for respective subbands of each frame based on the measured signal to mask ratio data and the generated bit allocation information, and bit allocation corresponding to the bits determined for the respective subbands. An adaptive digital audio encoding device having a second bit allocation means for generating information. The method of claim 2, wherein the decision levels (D) of the frame group are determined as follows. Where k (k = -q to q) is the decision level index, q is a positive integer, NF is the normalization factor in the frame group, and PEstd is the standard deviation parameter of the frame group. Encoding device. The method of claim 3, wherein the bit allocation (FB (p)) of the p-th frame in the frame group is calculated as follows: Where p is a frame index in the frame group, FBm is an average bit for one frame, BV is a predetermined bit shift value, 2q + 1 is the total number of predetermined decision levels, and I is the pth Adaptive digital audio encoding device that is a level index of a frame. A bit allocation method for adaptively encoding an input digital audio signal having a plurality of frames, wherein each frame includes a plurality of subbands, the method comprising: filtering the input digital audio signal to filter a plurality of subbands. Dividing into band units; Measuring signal-to-mask ratio data, sound pressure levels, and masking thresholds for respective subbands of the digital audio signal; Measure the amount of cognitive information for each frame of the input digital audio signal based on the measured signal-to-mask ratio data, sound pressure levels and masking thresholds, and at least two currents corresponding to the measured cognitive information amounts; Measuring average and standard deviation parameters for a frame group containing previous frames; Measuring decision levels of the frame group based on the measured mean and standard deviation parameters; Determine bits for each frame of the frame group based on the measured decision levels, the total number of decision levels, the amount of cognitive information, and a predetermined average bit, and the bits determined for each of the frames. Generating bit allocation information corresponding to the; And determine bits for respective subbands of each frame based on the measured signal to mask ratio data and the generated bit allocation information, and bits corresponding to the bits determined for the respective subbands. And generating allocation information.