High efficiency digital data encoding and decoding apparatus

• US 5,414,795 A
• Filed: 03/26/1992
• Issued: 05/09/1995
• Est. Priority Date: 03/29/1991
• Status: Expired due to Fees

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Tsutsui, Kyoya, Shimoyoshi, Osamu
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Doerrler, Michelle
• Application Number:
  US07/857,772
• Time in Patent Office: 1,139 Days
• Field of Search: 381/29-49, 395/2, 395/2.12, 395/2.13, 395/2.14, 395/2.38, 395/2.39
• US Class Current: 704/204
• CPC Class Codes: G10L 19/0212 using orthogonal transformation
• CPC Class Codes: G11B 20/00007 Time or data compression or expansion audio compression based on psychoacoustics G10L19/00; data processing for reproducing audio data at different playback speeds G10L21/04; video compression H04N19/00; data compression per se H03M7/30
• CPC Class Codes: G11B 20/10527 Audio or video recording; Data buffering arrangements G11B20/12 - G11B20/18 take precedence
• CPC Class Codes: H04B 1/665 using psychoacoustic properties of the ear, e.g. masking effect

Abstract

An apparatus for compressing a digital input signal compresses a digital input signal arranged into frames of plural samples. The digital input signal is orthogonally transformed by plural orthogonal transform circuits in blocks derived by dividing the frames by a different divisor in each circuit. The resulting spectral coefficients are quantized using an adaptive number of bits. The output of one orthogonal transform circuit is selected based on the outputs of the orthogonal transform circuits. The digital input signal is divided into plural frequency ranges, and frames are formed in each range. A block length decision circuit determines division of the frames of each range signal into blocks in response to range signal dynamics. The range signals are orthogonally transformed in blocks and the spectral coefficients are quantized. The apparatus also includes a frequency analyzing circuit that derives spectral data points from the input data, and groups them in plural bands. A noise level setting circuit finds the energy in each band and sets a allowable noise level in response to the band energy. A block floating coefficient calculating circuit calculates block floating coefficients based on the maximum spectral data point in each band. A quantizing bit number decision circuit determines the number of bits to use for quantizing in response to the allowable noise level and the block floating coefficients. A quantizing circuit quantizes the spectral data points in response to the quantizing bit number decision circuit.

Claims

1. An apparatus for compressing a digital input signal divided into frames of plural samples to provide a compressed signal, the apparatus comprising:

plural orthogonal transform means for orthogonally transforming the digital input signal in blocks derived by dividing each frame by a different set of divisors, each set of divisors including at least one divisor, each orthogonal transform means deriving spectral coefficients from the digital input signal, quantizing the spectral coefficients by adaptively allocating a number of quantizing bits among the spectral coefficients, each orthogonal transform means allocating an equal number of quantizing bits, and providing the quantized spectral coefficients as an output signal; and

means for selecting the output signal from one of the orthogonal transform means as the compressed signal in response to a selection signal from each of the orthogonal transform means.

2. An apparatus for compressing a digital input signal according to claim 1, wherein:

the apparatus additionally includes means for dividing the digital input signal in frequency into plural frequency range signals,

at least one of the orthogonal transform means transforms each frequency range signal, and

the orthogonal transform means transforming a frequency range signal towards higher frequencies transforms the frequency range signal in blocks derived by dividing by a greater divisor than orthogonal transform means transforming a frequency range signal towards lower frequencies.

3. An apparatus for compressing a digital input signal according to claim 1, wherein:

each orthogonal transform means quantizes the spectral coefficients using a primary number of bits and provides as the selection signal information indicating the primary number of bits; and

the selection means selects as the compressed signal the output signal from the orthogonal transform means providing the selection signal indicating the least primary number of bits.

4. A method according to claim 3, wherein:

in the step of quantizing the spectral coefficients, the spectral coefficients from each orthogonal transform are initially quantized using a primary number of bits, and information indicating the primary number of bits is provided as the selection signal; and

the step of selecting the output signal from one of the orthogonal transforms selects the output signal from the orthogonal transform providing the selection signal indicating the least primary number of bits.

5. An apparatus for compressing a digital input signal according to claims 1 or 2, wherein each orthogonal transform means performs a discrete cosine transform.

6. An apparatus for compressing a digital input signal according to claims 1 or 2, wherein each orthogonal transform means performs a fast Fourier transform.

9. An apparatus for compressing a digital input signal according to claim 4, wherein the block length decision means dynamically determines the division of each frame into at least one block such that the frames of the frequency range signals are divided into a number of blocks up to a maximum number of blocks, the maximum number of blocks for frequency range signals towards higher frequencies being greater than the maximum number of blocks for the frequency range signals towards lower frequencies.

11. An apparatus for compressing a digital input signal according to claims 3, 4, 5 or 6, wherein the orthogonal transform means performs a discrete cosine transform.

7. An apparatus for compressing a digital input signal to provide a compressed signal, the apparatus comprising:

means for dividing the digital input signal into plural frequency ranges to provide plural frequency range signals, each frequency range signal being arranged into frames composed of plural samples;

a block length decision means for dynamically determining a division of each frame of each frequency range signal into at least one block in response to a dynamic characteristic of the frame of the frequency range signal;

orthogonal transform means for orthogonally transforming each frequency range signal to produce spectral coefficients, the orthogonal transforms means transforming each frequency range signal in blocks determined by the block length decision means; and

means for quantizing the spectral coefficients.

8. An apparatus for compressing a digital input signal according to claim 7, wherein:

each frequency range has a bandwidth, and

the means for dividing the digital input signal into plural frequency ranges divides the digital input signal such that the bandwidths of the frequency ranges are wider towards higher frequencies.

10. An apparatus for compressing a digital input signal according to claim 8, wherein:

the frequency range signals include a middle frequency range signal and a higher frequency range signal, and

the block length decision means dynamically determines a division of the frames of the middle frequency range signal into blocks and independently dynamically determines a division of the frames of the high frequency range signal into blocks.

12. An apparatus for compressing a digital input signal according to claims 7, 8, 9 or 10, wherein the orthogonal transform means performs a fast Fourier transform.

17. An apparatus for compressing a digital input signal according to claims 9, 10 or 11, wherein the frequency analyzing means performs a fast Fourier transform.

13. A apparatus for compressing a digital input signal, apparatus comprising:

frequency analyzing means for frequency analyzing the digital input signal to provide plural spectral data points and for forming the spectral data points into plural bands;

allowable noise level setting means for determining an energy for each band, and for setting an allowable noise level in response to the energy in each band;

block floating coefficient calculating means for calculating a block floating coefficient in response to a maximum spectral data point in each band;

quantizing bit number decision means for determining a number of quantizing bits in response to the allowable noise level and the block floating coefficient; and

quantizing means for quantizing the spectral data points from the frequency analyzing means using the number of quantizing bits determined by the quantizing bit number decision means.

14. An apparatus for compressing a digital input signal according to claim 13, wherein the allowable noise level setting means:

is additionally for calculating a masking level from the energy in each band, and

sets the allowable noise level in response to the masking level.

15. An apparatus for compressing a digital input signal according to claim 14, wherein the allowable noise level setting means calculates the masking level taking account of an equal loudness characteristic of human auditory sense.

16. An apparatus for compressing a digital input signal according to claims 13, 14 or 15, wherein the frequency analyzing means performs a discrete cosine transform.

18. In combination, a compressor apparatus for compressing a digital input signal comprising plural samples to provide a compressed signal, and an expander apparatus for applying complementary expansion to the compressed signal to provide a digital output signal, the compressor comprising:

means for arranging the digital input signal in frames of plural samples,

means for dividing the digital input signal data into plural frequency ranges to provide plural frequency range signals,

means for orthogonally transforming each frequency range signal to provide spectral coefficients, each frequency range signal being transformed by dynamically dividing each frame of the frequency range signal into blocks having a block length determined in response to a dynamic characteristic of the frame of the frequency range signal, and

means for quantizing the spectral coefficients using an adaptive bit allocation,

means for including in the compressed signal the quantized spectral coefficients and subsidiary information indicating the adaptive bit allocation; and

the expander comprising;

means for extracting the quantized spectral coefficients and the subsidiary information from the compressed signal,

means for dequantizing the quantized spectral coefficients in response to the subsidiary information indicating the adaptive bit allocation to provide plural spectral coefficients,

means, including a reverse orthogonal transform circuit, for transforming the spectral coefficients into frequency range signals, and

means, including an inverse band division filter, for synthesizing the frequency range signals to produce the digital output signal.

19. A combination according to claim 18 wherein the reverse orthogonal transform circuit is a reverse discrete cosine transform circuit.

20. A combination according to claim 18 wherein each the inverse band division filter includes a reverse quadrature mirror filter circuit.

21. A method for storing in a medium a compressed digital signal derived from a digital input signal divided into frames of plural samples, the method comprising the steps of:

performing plural orthogonal transforms of the digital input signal each orthogonal transform;

orthogonally transforming the digital input signal in blocks derived by dividing a frame of the digital input signal by a different set of divisors, each set of divisors including at least one divisor to provide spectral coefficients,

quantizing the spectral coefficients by adaptively allocating a number of quantizing bits among the spectral coefficients using an adaptive bit allocation, each orthogonal transform adaptively allocating an equal number of quantizing bits and

providing the quantized spectral coefficients as an output signal for the orthogonal transform;

selecting, as the part of the compressed signal, the output signal for one of the orthogonal transforms in response to a selection signal for each orthogonal transform;

including subsidiary information indicating the adaptive bit allocation in the compressed signal; and

storing the compressed signal in the medium.

22. A method according to claim 21, wherein:

the method additionally includes the step of dividing the digital input signal in frequency into plural frequency range signals; and

in the step of performing plural orthogonal transforms;

at least one orthogonal transform is performed on each frequency range signal,

orthogonal transforms performed on frequency range signals towards higher frequencies are performed on blocks derived by dividing each frame by a greater divisor than orthogonal transforms performed on frequency range signals towards lower frequencies.

23. A method for storing in a medium a compressed digital signal derived from a digital input signal, the method comprising the steps of:

dividing the digital input signal into plural frequency ranges to provide plural frequency range signals, each frequency range signal being arranged into frames of plural samples;

dynamically determining a division of each frame of each frequency range signal into blocks in response to a dynamic characteristic of the frame of the frequency range signal;

orthogonally transforming each frequency range signal in blocks determined by the determining step to produce spectral coefficients;

quantizing the spectral coefficients using an adaptive bit allocation;

including the quantized spectral coefficients and subsidiary information indicating the adaptive bit allocation in the compressed digital signal; and

storing the compressed digital signal in the medium.

24. A method according to claim 23, wherein, in the step of dividing the digital input signal into plural frequency ranges, the digital input signal is divided into frequency ranges, each having a bandwidth, such that the bandwidths of the frequency ranges widen towards higher frequencies.

25. A method according to claim 24, wherein, in the step of determining a division of each frame of each frequency range signal into blocks, the frames of the frequency range signals are divided into a number of blocks up to a maximum number of blocks, the maximum number of blocks for frequency range signals towards higher frequencies being greater than the maximum number of blocks for frequency range signals towards lower frequencies.

26. A method for storing in a medium a compressed digital signal derived from a digital input signal, the method comprising the steps of:

frequency analyzing the digital input signal to provide plural spectral data points;

forming the spectral data points into plural bands;

determining an energy for each band;

setting an allowable noise level in response to the energy in each band;

calculating a block floating coefficient for each band in response to a maximum level spectral data point in the band;

determining a number of quantizing bits for each band in response to the allowable noise level and the block floating coefficient for the band;

quantizing the spectral data points in each band using the number of quantizing bits determined in the step of determining the number quantizing bits;

including the quantized spectral data points and subsidiary information indicating the number of quantizing bits for each band in the compressed digital signal; and

storing the compressed digital signal in the medium.

27. A method according to claim 26, wherein the step of setting the allowable noise level includes the steps of:

calculating a masking level from the energy in each band, and setting the allowable noise level in response to the masking level.

28. A method according to claim 27, wherein the step of calculating the masking level takes account of an equal loudness characteristic of human auditory sense.

29. A method according to claim 26, wherein the step of frequency analyzing the digital input signal includes the step of orthogonally transforming the digital input signal.