Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04B—TRANSMISSION
  • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04H—BROADCAST COMMUNICATION
  • H04H20/00—Arrangements for broadcast or for distribution combined with broadcast
  • H04H20/86—Arrangements characterised by the broadcast information itself
  • H04H20/95—Arrangements characterised by the broadcast information itself characterised by a specific format, e.g. an encoded audio stream
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
  • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
  • H03M7/40—Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code
  • H03M7/42—Conversion to or from variable length codes, e.g. Shannon-Fano code, Huffman code, Morse code using table look-up for the coding or decoding process, e.g. using read-only memory

Definitions

• the invention relates to a method for data reduction in the storage and / or transmission of digital audio signals according to the preamble of claim 1.
• Such data reduction methods are used when interference-free, high-quality sound is to be achieved when storing and / or transmitting audio signals, but the transmission or memory bandwidth is not as high as, for example, a compact disc (CD).
• CD compact disc
• Examples are the planned digital terrestrial broadcasting, the sound channel of digital video recorders or tape recorders with a stationary sound head and many more.
• a method according to the preamble has become known, for example, from DE-OS 3912605.
• the quantized spectral values are encoded with an optimal encoder, in which the code word used for encoding is shorter the more frequently the spectra coefficient occurs.
• the code is taken from a table, the length of which corresponds to the number of code Words corresponds to (Huffman code).
• the object of the invention is to develop a method for reducing the data according to the preamble such that it is suitable for professional applications in studios.
• the audio signal is mapped from the time range into the frequency range with a filter bank with perfect reconstruction. This means that without quantization and coding, the signal transmitted back into the time domain completely matches the input signal.
• a block section under consideration is coded with such a small number of bits that the noise caused by the quantization is still below the monitoring threshold, that is to say is covered by the music signal.
• the initially concealed quantization noise can become an audible disturbing noise.
• the invention at most half of the psychoacoustic concealment effect, compared to the aforementioned ASPEC method, is used in the quantization and coding. As a result, a distance between the concealed and concealing signal is achieved, which enables postprocessing of the music signal.
• the coding is carried out with a variable word length, a cascaded, multidimensional Huffman code being used as the code.
• the Huffman coding is thus carried out differently than with known methods (eg ASPEC).
• known methods predominantly have low values due to the high data reduction which have to be encoded.
• Code values are assigned to these values using a Huffman code table.
• the coding of very large quantized values, which are not included in the table, takes place with the help of an identifier, which initiates a further coding step.
• the coding is carried out using different Huffman code tables.
• the code word is assigned in accordance with this area by means of the suitable Huffman code table.
• the portion exceeding the threshold is coded scalarly. With small values, the coding is done in a known manner.
• the curve of the monitoring thresholds (English “spreading function") is used to calculate the masking effects.
• a curve with at least twice the slope is selected according to claim 2.
• the tendrils of the curve of the listening thresholds are preferably chosen to be so steep that only a slight concealment is assumed. With this coding, the quantization noise does not reach an audible range due to post-processing such as filtering for matching the signal.
• the amplitude of the noise resulting from the quantization is at least 14 dB below the masking audio signal. This ensures that there is no audible deterioration of the music signal even when the method is used several times.
• a level is assumed for the basic hearing threshold which is at least 10 dB below the amplitude resolution ("least significant bit").
• a modified discrete cosine transformation is used as the ferbank, which is used according to claim 6 with a block length of 256 or 512 samples. At a sampling frequency of 48 kHz, this results in block lengths of 5.33 and 10.67 msec.
• the MDCT is a transformation with perfect reconstruction, that is to say that the output signal of the synthesis filter bank is exactly the same as the input signal of the analysis filter bank if no quantization is carried out.
• the option to use a block length of 256 samples enables the time resolution to be increased while the coding efficiency is reduced.
• the analysis window contains 512 or 1024 samples and is therefore 10.67 or 21.33 msec long at a sampling rate of 48 kHz.
• the method also supports a sampling frequency of 96 kHz. At this sampling frequency, frequencies far above human perception can be transmitted with the method. This is sometimes called a requirement for studio applications.
• the encoded signal is then post-processed, for example hidden or shown.
• the coded signal can be entered into a computer and processed directly. With this method, great savings can be achieved in digital process processing and memory allocation. Since many music channels are usually mixed in a recording studio, up to 48 channels can be processed in parallel in a further development of the invention.
• An important advantage of the invention is that a digital music signal processed by the method is suitable for professional applications in recording studios.
• the signal can be mixed and faded, it can be played at very different volumes, it can be encoded and decoded several times in succession without a deterioration of the sound signal being perceptible.
• the signal can be provided with reverberation or other effects. All of the measures mentioned can be applied to individual channels which are combined to form an overall signal.
• the method according to the invention achieves a significantly greater reduction.
• a digital audio signal is mapped by a modified discrete cosine transformation from the time domain to the frequency domain in order to split the input signal into undersampled spectral components lay.
• the sampling frequency is 48 kHz
• the block length is 516 samples or 10.67 msec.
• the window width is 1024 samples.
• the output signals from this filter bank are used to calculate estimates of the respective signal-dependent monitoring thresholds.
• regularities known from psychoacoustics are used.
• a modified acoustic model is used to determine the minimum listening threshold.
• Model 2 of the encoder of the ISO / MPEG standard 11172-3 International Standardization Organization / Moving Pictures Experts Group
• the slope of the edges of the curve of the listening threshold is increased so much that only a minimal masking is taken into account.
• the use of a curve of the listening threshold with an incline of the flanks that has at least twice the slope of the curve of model 2 of the ISO / MPEG standard has proven to be particularly advantageous.
• the distance between the masking signal and the masked noise is 20 dB instead of 6-9 dB for noise-like signals.
• the signal is quantized using a linear quantizer because this enables an optimal ratio of masking signal and masked noise.
• the quantized spectral values are encoded using Huffman encoding. Since much more large values are to be encoded than, for example, in the known ASPEC method, the code tables are organized differently. Large values are encoded by determining the range in which the value falls. The corresponding area is then selected from various Huffman code tables. If the quantized values exceed a certain threshold, scalar values are coded instead of value pairs by the table.
• the coded values and the page information are combined to form an output bit stream. Since the psychoacoustic model is only used in the encoder and not in the decoder, the model can be changed without changing the definition of the bit stream.

Landscapes

• Engineering & Computer Science (AREA)
• Signal Processing (AREA)
• Theoretical Computer Science (AREA)
• Multimedia (AREA)
• Computer Networks & Wireless Communication (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)
• Signal Processing For Digital Recording And Reproducing (AREA)
• Circuits Of Receivers In General (AREA)
• Reduction Or Emphasis Of Bandwidth Of Signals (AREA)
• Transmission Systems Not Characterized By The Medium Used For Transmission (AREA)
• Stereo-Broadcasting Methods (AREA)

Priority Applications (8)

Applications Claiming Priority (2)

Publications (1)

Family

ID=6454791

Family Applications (1)

Country Status (9)

Families Citing this family (4)

Citations (4)

Family Cites Families (1)

• 1992
  • 1992-03-23 DE DE4209382A patent/DE4209382C1/de not_active Expired - Lifetime
• 1993
  • 1993-02-25 CA CA2131806A patent/CA2131806A1/en not_active Abandoned
  • 1993-02-25 JP JP5516152A patent/JPH07508375A/ja active Pending
  • 1993-02-25 AT AT93905148T patent/ATE197105T1/de active
  • 1993-02-25 WO PCT/DE1993/000177 patent/WO1993019536A1/de not_active Ceased
  • 1993-02-25 EP EP93905148A patent/EP0632940B1/de not_active Expired - Lifetime
  • 1993-02-25 DE DE59310119T patent/DE59310119D1/de not_active Expired - Lifetime
  • 1993-02-25 AU AU36254/93A patent/AU670068B2/en not_active Expired
• 1994
  • 1994-09-22 KR KR1019940703290A patent/KR950701161A/ko not_active Ceased
  • 1994-09-22 NO NO943520A patent/NO943520D0/no unknown

Patent Citations (4)

Also Published As

Similar Documents

Legal Events