Classifications

• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
  • G10L19/16—Vocoder architecture
  • G10L19/18—Vocoders using multiple modes
  • G10L19/24—Variable rate codecs, e.g. for generating different qualities using a scalable representation such as hierarchical encoding or layered encoding
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/008—Multichannel audio signal coding or decoding using interchannel correlation to reduce redundancy, e.g. joint-stereo, intensity-coding or matrixing
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04S—STEREOPHONIC SYSTEMS 
  • H04S1/00—Two-channel systems
  • H04S1/007—Two-channel systems in which the audio signals are in digital form
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
  • G10L19/0212—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders using orthogonal transformation

Definitions

• the present invention relates to scalable audio coders and in particular to methods and devices for coding a discrete-time stereo signal.
• Scalable audio encoders are encoders that have a modular structure. So there is an effort to use already existing speech coders, the signals which, for. B. are sampled at 8 kHz, process and output data rates of, for example, 4.8 to 8 kilobits per second.
• These known encoders such as. B. the encoders G. 729, G.723, FS1016, CELP or parametric models of the MPEG-4 audio VM, which are known to those skilled in the art, are mainly used for encoding voice signals and are generally not suitable for encoding higher quality music signals since they are usually designed for signals sampled at 8 kHz, which is why they can only encode an audio bandwidth of maximum 4 kHz. However, they generally show fast operation and little computing effort.
• a voice encoder is therefore combined with an audio encoder in a scalable encoder, which signals with a higher sampling rate, such as e.g. B. 48 kHz, can encode.
• a scalable encoder which signals with a higher sampling rate, such as e.g. B. 48 kHz, can encode.
• a speech coder it is also possible to replace the above-mentioned speech coder with another coder, for example with a music / audio coder according to the standards MPEG1, MPEG2 or MPEG4.
• a high quality audio encoder usually uses the method of differential coding in the time domain.
• An input signal which has a sampling rate of 48 kHz, for example, is down-sampled to the sampling frequency suitable for the speech coder by means of a downsampling filter.
• the down-sampled signal is now encoded.
• the coded signal can be fed directly to a bit stream formatting device in order to be transmitted. However, it only contains signals with a bandwidth of e.g. B. maximum 4 kHz.
• the coded signal is further decoded and sampled up using an upsampling filter. However, due to the downsampling filter, the signal now obtained only has useful information with a bandwidth of, for example, 4 kHz.
• the spectral content of the up-sampled coded / decoded signal in the lower band up to 4 kHz does not exactly correspond to the first 4 kHz band of the input signal sampled at 48 kHz, since encoders generally introduce coding errors.
• a scalable encoder has both a generally known speech encoder and an audio encoder that can process signals at higher sampling rates.
• a difference between the input signal at 8 kHz and the coded / decoded, sampled output signal of the speech encoder is formed for each individual time-discrete sample value.
• This difference can then be quantized and encoded using a known audio encoder, as is known to those skilled in the art.
• the difference signal which is fed into the audio encoder which can encode signals with higher sampling rates, is very much smaller in the lower frequency range, apart from coding errors of the speech encoder, than the original.
• the difference signal corresponds to essentially the true input signal, which with z. B. 48 kHz was sampled.
• a coder with a low sampling frequency is usually used, since generally a very low bit rate of the encoded signal is aimed for.
• coders also called coders
• They work with bit rates of a few kilobits (two to eight kilobits or even higher). They also allow a maximum sampling frequency of 8 kHz, since audio bandwidth at this low bit rate is no longer possible anyway, and coding at a lower sampling frequency is more economical in terms of computational complexity.
• the maximum possible audio bandwidth is 4 kHz and is limited in practice to about 3.5 kHz.
• “Joint stereo” includes stereo coding techniques, such as. B. to understand the middle-side coding (M / S coding) or the intensity stereo coding (IS coding). If a separate scalable mono audio encoder is simply used for the left (L) and the right (R) channel of a stereo signal, a stereo signal can be encoded, but the encoding does not take into account any joint stereo techniques, which can open up far-reaching savings opportunities in the bit-saving coding of stereo signals.
• M / S coding middle-side coding
• IS coding intensity stereo coding
• the object of the present invention is to provide a method and a device for coding a time to create discrete stereo signals that allow the use of joint stereo techniques.
• This object is achieved by a method for coding a time-discrete stereo signal according to claim 1 and by a device for coding a time-discrete stereo signal according to claim 14.
• the present invention is based on the finding that a combination of joint stereo techniques with the principle of scalability can be achieved if a mono signal is initially formed from the left and right channels of a stereo signal, which can preferably be done by summation.
• the mono signal is encoded by means of a first encoder, whereupon the resulting signal is fed to a bit stream multiplexer.
• the encoded mono signal is further decoded again to obtain an encoded / decoded mono signal that differs from the original mono signal in that it has coding errors introduced by the first encoder.
• stereo information can now be generated, which for example center / side (M / S) information or intensity stereo (IS) information or also under certain circumstances can be the original left channel or the original right channel.
• M / S center / side
• IS intensity stereo
• the coded / decoded mono signal itself or the difference of the original mono signal from the coded / decoded mono signal can also be used as stereo information, together with the difference between the left and right channel, which is also referred to as an S signal will result in a middle / side coding directly.
• the stereo information can now be coded by means of a second encoder, which can be constructed identically to the first encoder or also differently from the first encoder, and also fed to a bit stream multiplexer, which generates a bit stream from the encoded mono signal and the encoded stereo information and generated from page information necessary for later decoding.
• the formation of the mono signal and the coding of the same can take place in the time domain if, for example, as the first encoder or core encoder.
• the formation and coding of stereo information preferably takes place in the frequency domain, since powerful encoders can then be used which operate according to the psychoacoustic model.
• the left and right channels are transformed into the frequency domain before further processing, which means that a frequency domain encoder can also be used for coding the mono signal, which can code using the psychoacoustic model as distortion-free as possible.
• the mono signal formed from the summation of the left and right channels must first be converted to the lower sampling frequency, which is also called downsampling.
• the mono signal converted to the lower sampling frequency is now encoded and decoded again, the encoded / decoded mono signal also having the lower sampling frequency.
• the encoded / decoded mono signal In order to be able to be related to the higher-sampled left and right channel to form stereo information, the encoded / decoded mono signal must be converted back to the sampling frequency of the discrete-time stereo signal, which is also referred to as upsampling.
• MDCT modified discrete cosine transformation
• the first encoder is operated at the same sampling rate that the discrete-time stereo signal has, downsampling and upsampling can of course be dispensed with.
• FIG. 1 shows a scalable stereo encoder with mono signal formation and coding in the time domain and center / side coding in the frequency domain according to a first exemplary embodiment of the present invention
• FIG. 2A shows a scalable stereo coder with mono signal formation and coding in the time domain and L / R or M / S coding in the frequency domain according to a second exemplary embodiment
• FIG. 2B is a more detailed illustration of the scalable stereo encoder of FIG. 2A;
• FIG. 3 is an expanded illustration of the scalable stereo encoder shown in FIG. 2A according to a third embodiment of the present invention.
• FIG. 4 shows a scalable stereo encoder with mono signal formation in the time domain and optional L / R or M / S coding in the frequency domain.
• the scalable stereo encoder receives a discrete-time stereo signal, the first or left channel L and a second or right channel R comprises.
• a sum signal is formed from the stereo signal, preferably by sampling summation by means of a summer 102, which is then multiplied by a factor of 0.5 by means of a multiplier 104 in order to generate a mono signal in this exemplary embodiment which corresponds to that of the M / S Coding known center signal is identical.
• the mono signal at the output of the multiplier 104 is fed into a downsampling filter 106 in order to convert the sampling rate thereof to a preferably lower sampling rate, which enables the mono signal to be encoded by means of a time-domain encoder which is part of the core codec 108.
• the encoded mono signal is written together with corresponding side information in a bit stream multiplexer 110, which generates a bit stream at its output 112, which is a coded representation of the discrete-time stereo signal.
• the encoded mono signal is decoded again in order to be converted back to the first sampling rate by means of an upsampling filter 114, so that the encoded / decoded mono signal is related to the left and right channels for later formation of stereo information can be set.
• the discrete-time stereo signal could, for example, by means of a first sampling rate, e.g. B. 48 kHz, have been sampled.
• the downsampling filter 106 could change this signal with the first sampling rate to a second sampling rate of e.g. B. implement 8 kHz.
• the first and the second sampling rate preferably form an integer ratio.
• the downsampling filter 106 can be implemented, for example, as a decimation filter.
• the core codec 108 could, for example, be a speech coder, e.g. BG729, G.723, FS1016, MPEG-4 CELP, MPEG-4 PAR, or a similar encoder.
• Such encoders operate at data rates from 4.8 kilobits per second (FS1016) to data rates from 8 kilobits per second (G.729).
• FS1016 4.8 kilobits per second
• G.729 8 kilobits per second
• the encoded mono signal has a maximum bandwidth of 4 kHz, since the downsampling filter 106 converts the mono signal z. B. has implemented by means of decimation to a sampling frequency of 8 kHz. Within the bandwidth of 0-4 kHz, the coded / decoded mono signal and the original mono signal at the input of the downsampling filter 106 are now the same, apart from coding errors introduced by the core codec 108.
• the coding errors introduced by the core codec 108 are not always small errors, but that they can easily come in the order of magnitude of the useful signal if, for example, a strongly transient signal is encoded in the first encoder. For this reason, as will be discussed later, it is checked whether differential coding makes sense at all.
• the output signal of the upsampling filter 114 is now converted into the frequency range by means of MDCT filter banks 116.
• the output signals of the MDCT filter banks 116 are fed directly to a first frequency-selective switching device (FSS) 118a or to a second frequency-selective switching device 118b or indirectly via a first summer 120a or a second summer 120b .
• FSS frequency-selective switching device
• the output signal of the MDCT filter bank for the left channel is fed to the first frequency-selective switching device (FSS) 118a, which also receives the sum of the transformed left channel and the transformed coded / decoded mono signal provided with a negative sign.
• the second frequency-selective switching device 118b receives the Sum of the transformed R-channel and the coded / decoded mono signal with negative sign.
• the frequency-selective switching devices 118a, 118b check whether it is more favorable to process the transformed original left or right " signal or the difference between the left or right signal and the encoded / decoded mono signal. The function of the frequency-selective switching device will be described in more detail later .
• the output signal of the first frequency-selective switching device 118a is fed to both a third summer 122a and a fourth summer 122b with a positive sign, while the output signal of the second frequency-selective switching device 118b is fed to the third summer 122a with a positive sign and the fourth summer 122b with a negative sign.
• the third summer 122a there is either either the sum of the transformed left and right channel or the difference between the sum of the uncoded left and right channel and the coded / decoded sum of the left and right channel.
• This signal which now has stereo information in contrast to the encoded mono signal of the core codec 108, is encoded by means of an M encoder 124, taking into account the psychoacoustic model, for example, and fed to the bitstream multiplexer 110.
• the fourth summer 122b there is the difference of the transformed left and right channels, this signal also being referred to in the art as a side signal which is fed into an S-encoder 126, the S-encoder 126 as well the M encoder 124 can encode considering the psychoacoustic model.
• the output signal of the S encoder 126 is likewise fed into the bit stream multiplexer and also comprises stereo information with regard to the discrete-time stereo signal at the input of the scalable stereo encoder 100 according to the first exemplary embodiment of the present Invention. It will be apparent to those skilled in the art that a complete bit stream requires page information.
• Side information relevant to the invention is, in particular, information from the frequency-selective switching devices 118a and 118b with regard to the frequency band in which differential signals or transformed L or R signals were output to the third summer 122a or the fourth summer 122b.
• the output signal of the core codec 108 has, as already mentioned, e.g. B. a sampling frequency of 8 kHz.
• This signal i.e. H. the mono signal, with a lower sampling rate than the original discrete-time stereo signal, is now to be related to the left or right channel in order to form stereo information.
• the signal with a lower sampling rate must therefore be converted into a signal with the same sampling rate as the sampling rate of the discrete-time stereo signal.
• the number of zero values is calculated from the ratio of the first and the second sampling frequency.
• the ratio of the first (high) to the second (low) sampling frequency is called the upsampling factor.
• the insertion of zeros which is possible with very little computational effort, produces an aliasing disturbance which has the effect that the low-frequency or zero spectrum of the encoded / decoded mono signal is repeated at the output of the core codec 108 is, in total, as often as many zeros have been inserted.
• the aliasing signal is now in the frequency range by means of the MDCT filter bank 116 transformed.
• the coded / decoded mono signal converted up to the first sampling frequency is only a correct representation of the original mono signal at the output of the multiplier 104 in the lower frequency band, which is why only a maximum of one / upsampling factor times the total spectral lines is used at the output of the MDCT filter bank 116 becomes.
• inserting the zeros into the encoded / decoded mono signal at the output of core codec 108 causes the spectral representation of the encoded / decoded mono signal to now have the same time and frequency resolution as the transformed left and right channels.
• the frequency-selective switching devices therefore carry out a so-called simulcast difference switchover. For example, it is unfavorable to process a difference signal further if the difference signal has a higher energy than the corresponding other signal at the input of the frequency-selective switching device 118a. Since an arbitrary encoder can be used as the core codec 108, it can happen that the encoder produces certain signal components that are difficult to encode by the M-encoder 124 or by the S-encoder 126.
• the core codec 108 should preferably preserve phase information of the signal coded by it, which is referred to in the art as "waveform coding" or “signal form coding”.
• wave coding or “signal form coding”.
• the decision that the frequency-selective switching module 118a or 118b makes is preferably made as a function of frequency.
• “Differential coding” means that only the difference between the transformed left and right channels and the transformed coded / decoded mono signal is coded. However, if this differential coding is not favorable, since the energy content of the differential signal is greater than the energy content of the transformed left or right signal, then differential coding is dispensed with and a switch is made to simulcast operation.
• Forming stereo information based on the encoded / decoded mono signal and the first and second Channel therefore includes a determination of where it is more convenient to process the transformed left or right channel or a difference thereof and the encoded / decoded mono signal.
• a frequency-selective comparison of the respective energies is now carried out in each selected frequency band. If the energy in a particular frequency band of the difference signal exceeds the energy of the other signal multiplied by a predetermined factor k, it is determined that the output signal of the frequency selective switching device 118a is the original transformed left signal. Otherwise, it is determined that the difference spectral values are output.
• the factor k can range, for example, from about 0.1 to 10.
• simulcast coding is already used if the difference signal has a lower energy than the other signal.
• differential coding is still used, even if the energy content of the differential signal is already greater than that of the original left or right channel.
• formation of stereo information can also be carried out in such a way that, for. B. a ratio or other linkage of the encoded / decoded mono signal and the transformed left or right channel is implemented.
• FIG. 2A shows a scalable stereo encoder 200 according to a second exemplary embodiment of the present invention.
• the same elements have the same reference symbols and, if they behave in the same way, are not described again.
• the scalable stereo encoder 200 differs from the scalable stereo encoder 100 according to the first exemplary embodiment of the present invention essentially in that either a middle / side coding or an L / R coding can be carried out.
• the scalable stereo encoder 200 comprises further summing devices 202a, 202b in order to use the transformed left and right channel to generate a center signal M and a side signal S, respectively.
• the transformed encoded / decoded mono signal is referred to here as M '.
• the signal M and the signal M ' are fed into a likewise additional frequency-selective switching device 204, which generates a signal M'', the frequency-selective switching device 204 also being preceded by a summer 206, as is the case with all other frequency-selective switching devices.
• the scalable stereo encoder 200 also includes a block joint stereo decision 208 which receives 4 input signals L ', M'', S and R'. The block joint stereo decision 208 decides in a known manner whether L / R, M / S or intensity coding is to be carried out by a stereo encoder 210.
• a mono signal is formed from the discrete-time stereo signal, this formation taking place in the time domain and equatingly as follows:
• the index T is intended to indicate that this is a middle signal in the time domain.
• the core encoder 108 now operates as was shown in connection with FIG. 1.
• an MDCT is also carried out on the L and R signals.
• the M / S signal in the frequency domain is now calculated using the summers 202a and 202b and the downstream multipliers, which is expressed as follows in equations:
• the frequency-selective switching device now serves to calculate M “.
• M is either equal to M-M "or M itself, as has already been shown.
• the frequency-selective switching device 118 calculates the signal L ', which either is equal to 0.5 • (L - M ') or equal to 0.5 • L.
• the switching devices 118a, 118b and 204 operate in a frequency-selective manner.
• a decision is now made in the usual way as to whether the signals L 'and R' or M '' or S have to be encoded known in the art and is therefore not explained in detail.
• FIG. 2B shows a scalable stereo encoder, which differs in some points from the scalable stereo encoder 200 according to the second exemplary embodiment of the invention.
• the same comprises the two multipliers 214a and 214b, which are arranged after the frequency-selective switching device 204 and after the frequency-selective switching device 118b.
• 2B also includes a somewhat more detailed illustration of the frequency selective switching devices.
• the switch state of the frequency selective switch 118a which is referred to as S 1LR
• S ' 1LR will always be complementary to the switch state of the frequency selective switch 118b, which is referred to as S ' 1LR .
• S 2 assumes a different state, ie state b, as shown in the drawing, it is sufficient to transmit the state S 1M of the frequency-selective switching device 204 , which indicates whether a differential or simulcast coding of the signal M is carried out. If the switch S 2 is in a position c, then side information is transmitted that an intensity stereo coding is present, in which case the position is likewise of the switch S 1M is transmitted, while here the positions of S 1LR and S ' 1LR are irrelevant.
• FIG. 3 comprises a further exemplary embodiment 300 of a scalable stereo encoder according to the present invention.
• the embodiment shown in FIG. 3 differs from the embodiment shown in FIG. 2 essentially in that the mono signal is encoded in two stages.
• the first stage is formed by the core codec 108, while the second stage is formed by an encoder / decoder 302 which, in the preferred embodiment, operates in the frequency domain and can be designed as a psychoacoustic frequency domain encoder. It receives the output signal of the frequency-selective switching device 204 as the input signal M ′′, it also being checked here whether a differential or simulcast coding is useful or not.
• the output signal of the encoder / decoder 302 is fed to a summer 304, whose output signal M ′′ ′′ corresponds to the difference between the signal M and the output signal of the encoder / decoder 302. Like the signals L ', S and R', this signal M '''is (not shown) and then fed to a stereo encoder (also not shown).
• the core codec 108 like the encoder / decoder 302, includes an output to the bitstream multiplexer to transmit encoded data thereto.
• the bit stream comprises, in addition to the first layer or the first layer, which is formed by the encoded mono signal of the core codec 108, a second layer, which is encoded by the encoded signal M "at the bit current multiplexer output of the encoder / decoder 302 is formed, wherein the encoder 300 shown in FIG. 3 can enable encoding of the mono signal at full sampling rate.
• FIG. 4 shows a scalable audio encoder 400 which only carries out mono signal formation in the frequency domain.
• the signals L and R are transformed into the frequency range by means of MDCT filter banks 116, after which an M / S matrix is carried out by means of the summers 202a and 202b and the subsequent multipliers by a factor of 0.5.
• an M / S matrix is carried out by means of the summers 202a and 202b and the subsequent multipliers by a factor of 0.5.
• the middle signal which can be used as a mono signal, is encoded and decoded again by means of a first encoder / decoder 402, the encoded mono signal M being written into the bit stream, as has already been mentioned several times.
• a summation device 404 Downstream of the encoder / decoder 402 is a summation device 404 which forms the difference between the encoded / decoded mono signal " and the original mono signal M, this difference being designated M '.
• the signals L', M ', S and R' can again be fed to a joint stereo decision device, which, however, is not shown in FIG. 4.
• the encoder 400 presented in FIG. 4 thus operates completely in the frequency domain, the encoder / decoder 402 preferably being designed as a frequency domain encoder with a full sampling rate.
• the stereo encoder (not shown) after the IS decision stage (also not shown in FIG. 4) is preferably also designed as a frequency domain encoder with a full sampling rate.
• the scalable stereo encoder shown in FIG. 4 thus represents a generalization of the term "scalability", since the bit stream here does not include layers or "layers" with different audio bandwidths, but (like the other exemplary embodiments) comprises a monolayer and a stereo layer, which by a Encoders can be encoded separately.
• An older mono decoder which is not equipped for stereo operation, can thus, for example, decode the bit stream of the coders according to the invention in order to generate at least one mono audio signal.
• the scalable stereo encoders according to the invention are thus backwards compatible with existing monodecoders.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Multimedia (AREA)
• Signal Processing (AREA)
• Acoustics & Sound (AREA)
• Health & Medical Sciences (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Human Computer Interaction (AREA)
• Computational Linguistics (AREA)
• Quality & Reliability (AREA)
• Mathematical Physics (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)
• Stereo-Broadcasting Methods (AREA)
• Reduction Or Emphasis Of Bandwidth Of Signals (AREA)
• Transmission Systems Not Characterized By The Medium Used For Transmission (AREA)

Applications Claiming Priority (3)

Publications (2)

Family

ID=7843796

Family Applications (1)

Country Status (7)

Families Citing this family (22)

Family Cites Families (11)

• 1997
  • 1997-09-26 DE DE19742655A patent/DE19742655C2/de not_active Expired - Lifetime
• 1998
  • 1998-06-15 AT AT98932156T patent/ATE205041T1/de active
  • 1998-06-15 DE DE59801343T patent/DE59801343D1/de not_active Expired - Lifetime
  • 1998-06-15 EP EP98932156A patent/EP1016319B1/fr not_active Expired - Lifetime
  • 1998-06-15 ES ES98932156T patent/ES2161059T3/es not_active Expired - Lifetime
  • 1998-06-15 US US09/445,894 patent/US6629078B1/en not_active Expired - Lifetime
  • 1998-06-15 WO PCT/EP1998/003605 patent/WO1999017587A1/fr active IP Right Grant
  • 1998-06-15 DK DK98932156T patent/DK1016319T3/da active

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events