RU2574849C2 - Apparatus and method for encoding and decoding audio signal using aligned look-ahead portion - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: windower for applying a prediction coding analysis window to a stream of audio samples to obtain windowed data for prediction analysis and for applying a transform coding analysis window to the stream of audio samples to obtain windowed data for transform analysis. The transform coding analysis window is associated with audio samples within a current frame of audio samples and with audio samples of a predefined portion of a future frame of audio samples which is a transform-coding look-ahead portion. The prediction coding analysis window is associated with the portion of the audio samples of the current frame and with audio samples of a predefined portion of the future frame which is a prediction coding look-ahead portion.

EFFECT: high quality of the encoded audio signal.

26 cl, 17 dwg

Description

The present invention relates to encoding an audio signal and, in particular, to encoding an audio signal based on switchable audio encoders and accordingly controlled audio decoders, especially suitable for low-latency applications.

Several audio coding concepts based on switchable codecs are known. One well-known audio coding concept is the so-called Advanced Broadband Adaptive Variable Speed Encoding (AMR-WB +) codec, as described in 3 GPP TS 26.290 B10.0.0 (2011-03). AMR-WB + audio codec contains all modes 1-9 of the AMR-WB codec for speech signals and AMR-WB VAD and DTX. AMR-WB + extends the AMR-WB codec by adding TCX, bandwidth extension and stereo.

AMR-WB + audio codec processes input frames equal to 2048 discrete values, with an internal sampling frequency F _S. The internal sampling rate is limited to the range of 12800-38400 Hz. Frames of 2048 discrete values are divided into two critically sampled equal frequency ranges. The result of this is two superframes of 1024 discrete values corresponding to the low-frequency (LF) and high-frequency (HF) ranges. Each superframe is divided into four frames of 256 discrete values. Sampling with an internal sampling rate is obtained by using an adjustable sampling conversion circuit that resambles the input signal.

LF and HF signals are then encoded using two different approaches: LF is encoded and decoded using a “main” encoder / decoder based on switchable ACELP and encoded transform excitation (TCX). In ACELP mode, the standard AMR-WB codec is used. The HF signal is encoded with a relatively small number of bits (16 bits / frame) using the bandwidth extension method (BWE). The parameters transmitted from the encoder to the decoder are mode selection bits, LF parameters, and HF parameters. The parameters for each superframe of 1024 discrete values are divided into four packets of the same size. When the input signal is a stereo signal, the left and right channels are combined into a mono signal for ACELP / TCX encoding, while stereo encoding receives both input channels. On the decoder side, LF and HF ranges are decoded separately, after which they are combined in a synthesizing filter bank. If the output signal is limited only to a mono signal, the stereo parameters are omitted, and the decoder operates in mono mode. The AMR-WB + codec uses LP analysis for both ACELP and TCX modes when encoding an LF signal. LP coefficients are linearly interpolated in each subframe with 64 discrete values. The LP analysis window is a half-sine with a length of 384 discrete values. For encoding the main mono signal for each frame, either ACELP or TCX encoding is used. The encoding mode is selected based on the feedback analysis method. For ACELP frames, only frames with 256 discrete values are considered, while in TCX mode, frames with 256, 512, or 1024 discrete values are possible. The window used for LPC analysis in AMR-WB + is depicted in FIG. 5b. A symmetric LPC analysis window is used with a 20 ms lead time. Advance viewing means that, as shown in FIG. 5b, the LPC analysis window for the current frame, shown at 500, extends not only within the current frame, designated between 0 and 20 ms in FIG. 5b and shown at 502, but extends into a future frame between 20 and 40 ms. This means that when using this LPC analysis window, an additional delay of 20 ms is required, i.e. an entire future frame. Therefore, the look-ahead portion designated 504 in FIG. 5b contributes to the systematic delay associated with the AMR-WB + encoder. In other words, the future frame must be fully accessible so that LPC analysis coefficients for the current frame 502 can be calculated.

FIG. 5a shows another encoder, the so-called AMR-WB encoder, and in particular, the LPC analysis window used to calculate the analysis coefficients for the current frame. Once again, the current frame extends between 0 and 20 ms, and the future frame extends between 20 and 40 ms. In contrast to FIG. 5b, the AMR-WB LPC analysis window, designated as 506, has a lead-in view portion 508 of only 5 ms, i.e., a time interval between 20 ms and 25 ms. Therefore, the delay introduced by the LPC analysis is significantly reduced relative to FIG. 5a. On the other hand, however, it was found that most of the leading look for determining the LPC coefficients, that is, most of the leading look for the LPC analysis window, leads to better LPC coefficients and therefore less energy in the residual signal and therefore lower bit rate, since LPC prediction matches better with the original signal.

Although FIG. 5a and 5b relate to encoders having only one analysis window for determining LPC coefficients for one frame, FIG. 5c depicts a situation for a G.718 speech encoder. Specification G718 (06-2008) relates to transmission systems and digital media systems and networks and, in particular, describes digital terminal equipment and, in particular, encoding voice and audio signals for such equipment. In particular, this standard refers to reliable narrow-band and wide-band embedded coding with variable speech and sound speeds from 8-32 Kbps, as defined in ITU-T G718. The input signal is processed using 20 ms frames. Codec delay depends on the sampling rate of the input and output. For broadband input and broadband output, the total algorithmic delay of this encoding is 42.875 ms. It consists of one 20 ms frame, 1.875 ms delay of the input and output resampling filters, 10 ms for the advanced viewing of the encoder, one ms post-filtering delay, and 10 ms in the decoder to provide encoding with conversion of higher layers. For narrow-band input and narrow-band output, higher layers are not used, but the 10 ms decoder delay is used to improve coding efficiency in the presence of frame erasure and for music signals. If the output is limited to layer 2, the codec delay can be reduced by 10 ms. The description of the encoder is as follows. The two lower layers are applied to the pre-emphasis signal sampled at a frequency of 12.8 kHz, and the three upper layers are used in the area of the input signal sampled at a frequency of 16 kHz. The base layer is based on Code Excited Linear Prediction (CELP) technology, in which a speech signal is modeled by an excitation signal that passes through a linear prediction synthesis filter (LP), which is an envelope of the spectrum. The LP filter is quantized in the spectral frequency domain of the immitance (ISF) using a switchable predictive approach and multi-stage vector quantization. Feedback analysis of the pitch is performed by a pitch tracking algorithm to ensure a smooth pitch contour. The two simultaneous contours of the evolution of the fundamental tone are compared and a track is selected that produces a smoother outline to make the assessment of the fundamental tone more reliable. Pre-processing at the frame level includes high-pass filtering, converting the sampling frequency to 12800 discrete values per second, predistorting, spectral analysis, detecting narrow-band inputs, detecting voice activity, noise estimation, noise reduction, linear prediction analysis, converting LP to ISF and interpolating, calculation of the weighted speech signal, analysis of the fundamental tone without feedback, updating the background noise, classification of signals for choosing the encoding mode and mask vanie frame erasure. Encoding a layer 1 using a selected encoding type comprises an unvoiced encoding mode, a voiced encoding mode, a transient encoding mode, a general encoding mode, intermittent transmission, and comfort noise generation (DTX / CNG).

Analysis with long-term prediction or linear prediction (LP) using the autocorrelation approach determines the coefficients of the synthesis filter of the CELP model. In CELP, however, long-term prediction is usually an “adaptive codebook” and thus differs from linear prediction. Therefore, linear prediction can be considered more as a short-term prediction. Autocorrelation of speech processed using a window function is converted to LP coefficients using the Levinson-Darbin algorithm. The LPC coefficients are then converted to immitance spectral pairs (ISP) and, therefore, to immitance spectral frequencies (ISF) for interpolation and quantization. The interpolated quantized and inverse quantized coefficients are converted back to the LP region to construct synthesizing and weighting filters for each subframe. In the case of encoding an active signal frame, two sets of LP coefficients in each frame are estimated using two LPC analysis windows, indicated as 510 and 512 in FIG. 5c. Window 512 is called the "mid-frame LPC window", and window 510 is called the "end-of-frame LPC window". The 10 ms look-ahead portion 514 is used to calculate end-of-frame autocorrelation. The frame structure is shown in FIG. 5c. The frame is divided into four subframes, each subframe has a duration of 5 ms, which corresponds to 64 discrete values at a sampling frequency of 12.8 kHz. Windows for analyzing the end of the frame and for analyzing the middle of the frame are centered in the fourth subframe and second subframe, respectively, as shown in FIG. 5c. A Hamming window with a length of 320 discrete values is used for processing using a window function. Odds are defined in G.718, section 6.4.1. Autocorrelation calculation is described in section 6.4.2. The Levinson-Darbin algorithm is described in section 6.4.3, the conversion of LP to ISP is described in section 6.4.4, and the conversion of ISP to LP is described in section 6.4.5.

Speech encoding parameters such as delay and gain of the adaptive codebook, index and gain of the algebraic codebook are sought by minimizing the error between the input signal and the synthesized signal in a perceptually weighted region. Perceptual weighting is performed by passing a signal through a perceptual weighting filter obtained from the LP filter coefficients. A perceptually weighted signal is also used in open-loop pitch analysis.

The G.718 encoder is an exclusively speech encoder having only one speech encoding mode. Therefore, the G.718 encoder is not a switchable encoder and, therefore, this encoder has the disadvantage that it provides only one speech encoding mode within the main layer. Therefore, there will be quality problems when this encoder is applied to signals other than speech signals, that is, ordinary audio signals for which the model behind the CELP encoding is not suitable.

An additional switchable codec is the so-called USAC codec, that is, a unified codec for speech and sound, as defined in ISO / IEC CD 23003-3 of September 24, 2010. The LPC analysis window used for this switchable codec is indicated in FIG. 5d is numbered 516. Again, it is assumed that the current frame extends between 0 and 20 ms, and therefore it turns out that the forward look part 618 of this codec is 20 ms, that is, significantly larger than the G.718 look-ahead part. Therefore, although the USAC encoder provides good sound quality due to having a switch, the delay is significant due to the LPC analysis window portion 518 of the LPC analysis window of FIG. 5d. The general structure of USAC is as follows. Firstly, there is a general pre / post-processing, consisting of an MPEG environment function block (MPEGS) for stereo or multi-channel processing and an advanced SBR (eSBR) block that processes the parametric representation of higher audio frequencies in the input signal. In addition, there are two branches, one consisting of a modified advanced audio coding (AAC) instrument path, and the other consisting of a linear prediction coding path (LP or LPC region), which in turn shows or representation in the frequency domain, or representation in the time domain of the remainder of the LPC. All transmitted spectra for both AAC and LPC are presented in the MDCT domain after quantization and arithmetic coding. The time domain representation uses an ACELP excitation coding scheme. The ACELP tool provides a way to efficiently present an excitation signal in the time domain by combining a long-term predictive mechanism (adaptive codeword) with a pulse-like sequence (innovative codeword). The reconstructed excitation is sent through an LP synthesis filter to generate a signal in the time domain. The input for the ACELP tool contains adaptive and innovative codebooks, gain values of adaptive and innovative codes, other control data and inversely quantized and interpolated LPC filter coefficients. The output of the ACELP tool is a time-reconstructed audio signal.

The MDCT-based decoding tool TCX is used to return a weighted representation of the LP remainder from the MDCT region back to a signal in the time domain; it outputs a weighted signal in the time domain including the weighted synthesizing LP filtering. IMDCT can be configured to support 256, 512, or 1024 spectral coefficients. The input for the TCX instrument contains (back quantized) MDCT spectra and back quantized and interpolated LPC filter coefficients. The output of the TCX instrument is a reconstructed time-domain audio signal.

FIG. 6 shows a situation in USAC, where LPC analysis windows 516 for the current frame and 520 for past or last frames are shown, and where, in addition, TCX window 522 is shown. Window 522 TCX has a center in the center of the current frame, extending between 0 and 20 ms and extending for 10 ms in the last frame and 10 ms in the future frame, extending between 20 and 40 ms. Therefore, the LPC analysis window 516 requires a LPC look-ahead portion between 20 and 40 ms, i.e. 20 ms, while the TCX analysis window further has a look-ahead portion extending between 20 and 30 ms in a future frame. This means that the delay introduced by the USAC analysis window 516 is 20 ms, while the delay introduced by the TCX window into the encoder is 10 ms. Therefore, it becomes clear that parts of the look-ahead of both types of windows are not aligned with each other. Therefore, even though the TCX window 522 introduces only a delay of 10 ms, the overall encoder delay is nevertheless equal to 20 ms due to the LPC analysis window 516. Therefore, even though there is a rather small part of the look-ahead for the TCX window, this does not reduce the overall algorithmic delay of the encoder, since the total delay is determined by the largest contribution, i.e., it is equal to 20 ms because of the LPC analysis window 516, which extends for 20 ms in future frame, that is, covering not only the current frame, but also further covering the future frame.

It is an object of the present invention to provide an improved coding concept for encoding or decoding an audio signal, which, on the one hand, provides good audio quality and which, on the other hand, leads to a reduction in delay.

This goal is achieved using a device for encoding an audio signal according to claim 1, a method of encoding an audio signal according to claim 15, an audio decoder according to claim 16, a method of decoding an audio signal according to claim 24, or a computer program according to claim 25.

In accordance with the present invention, a switchable audio codec circuit having a transform encoding branch and a predictive encoding branch is applied. It is important to note that two types of windows, that is, a predictive encoding analysis window on the one hand and a transform encoding analysis window on the other hand, are aligned with respect to their part of the leading look-up so that the part of the leading look-up of the encoding with conversion and the part of the look-ahead of the encoding with less than 20% of the predictive coding preview portion of the transform are identical or different from each other with prediction, or less than 20% of the predictive encoding preview portion of the transform. It should be noted that the forward analysis window is used not only in the prediction coding branch, but it is actually used in both branches. LPC analysis is also used to generate noise in the transform domain. Therefore, in other words, the parts of the leading look are identical or very close to each other. This ensures that an optimal compromise is achieved and that neither sound quality nor delay parameters are set in a suboptimal way. Therefore, for predictive coding, it was found in the analysis window that the LPC analysis is better, the larger the look-ahead, but, on the other hand, the delay increases for most look-ahead. On the other hand, the same is true for the TCX window. The larger the portion of the TCX window look-ahead, the stronger the TCX bit rate can be reduced, since longer TCX windows result in lower bit rates in general. Therefore, in accordance with the present invention, the parts of the leading viewing are identical or very close to each other and, in particular, differ from each other by less than 20%. Therefore, a portion of the look-ahead that is undesirable for delay reasons, on the other hand, is optimally used by both encoding / decoding branches.

In view of this, the present invention provides an improved coding concept, on the one hand, with a low latency, when the part of the look-ahead for both analysis windows is set small, and provides, on the other hand, a good coding / decoding concept due to the delay, which should be introduced for sound quality reasons or bit rate reasons, in any case it is optimally used by both coding branches, and not just by one coding branch.

An apparatus for encoding an audio signal having a discrete value stream comprises a window function application unit for applying a predictive coding analysis window to a stream of discrete audio values to obtain predictive analysis data processed by a window function and for applying a transform analysis coding window to a discrete value stream audio signal for receiving processed using the window function data for analysis with conversion. The transform coding analysis window is associated with discrete audio values of the current frame of discrete audio values of a given portion of the forward look-ahead of the future frame of discrete audio values that are part of the look-ahead of the transform coding.

In addition, the prediction encoding analysis window is associated with at least a portion of the discrete audio values of the current frame and with discrete audio values of a predetermined portion of the future frame that is part of the predictive encoding look-ahead.

The conversion coding look-ahead part and the predictive coding look-ahead part are identical to each other or less than 20% of the predictive coding look-ahead part or less than 20% of the look-ahead coding look-up and are therefore quite close to each other. The device further comprises a coding processor for generating predictive encoded data for the current frame using windowed data for advanced analysis, or for generating transform encoded data for the current frame using window data for transform analysis.

An audio decoder for decoding an encoded audio signal comprises a prediction parameter decoder for decoding data for a prediction encoded frame from the encoded audio signal and, for a second branch, a transform parameter decoder for decoding data for the encoded conversion frame from the encoded audio signal.

The transform parameter decoder is configured to perform a spectral-temporal transform, which is preferably a distorted transform, such as MDCT or MDST or any other similar transform, and to apply the synthesis window to the transformed data to obtain data for the current frame and future frame. The synthesis window used by the audio decoder is such that it has a first overlapping part, an adjacent second non-overlapping part and an adjacent third overlapping part, while the third overlapping part is associated with discrete audio signal values for the future frame, and the non-overlapping part is associated with the data of the current frame. In addition, in order to have good sound quality on the side of the decoder, an overlapping block is used to superimpose and add discrete values processed by the synthesis window associated with the third overlapping part of the synthesis window for the current frame and processed by the synthesis window of discrete values associated with the first overlapping part of the synthesis window for the future frame, to obtain the first part of the discrete values of the audio signal for the future frame, while the remaining discrete values of the audio signal for future frames are discrete values processed using the synthesis window associated with the second non-overlapping part of the synthesis window for the future frame obtained without adding overlap when the current frame and the future frame contain data encoded with conversion.

Preferred embodiments of the present invention have the feature that the leading views for the transform coding branch, such as the TCX branch, and the predictive coding branches, such as the ACELP branch, are identical to each other, so that both encoding modes have the most accessible leading viewing with limitations superimposed on the delay. In addition, it is preferable that the TCX window overlap be limited to a portion of the look-ahead, so that switching from a coding mode with conversion to a coding mode predicting from one frame to the next frame is easily possible without any distortion problems.

An additional reason to limit the overlap to leading viewing is the desire not to introduce a delay on the side of the decoder. If there were a TCX window with a 10 ms look-ahead and, for example, 20 ms overlap, this would introduce another 10 ms of delay in the decoder. When there is a TCX window with 10 ms look-ahead and 10 ms overlap, there is no additional delay on the decoder side. A good consequence of this is an easier switchover.

Therefore, it is preferable that the second non-overlapping part of the analysis window and, of course, the synthesis window extend to the end of the current frame, and the third overlapping part starts only relative to the future frame. In addition, the nonzero part of the TCX coding analysis / synthesis window or transform coding window is aligned with the beginning of the frame so that, again, it is possible to switch easily from one mode to another mode with low efficiency.

Furthermore, it is preferable that an entire frame consisting of a plurality of subframes, for example four subframes, can be either fully encoded in transform coding mode (such as TCX mode) or fully encoded in predictive coding mode (such as ACELP mode).

In addition, it is preferable to use not only one LPC analysis window, but two different LPC analysis windows, where one LPC analysis window is aligned with the center of the fourth subframe and is an end-of-frame analysis window, while the other analysis window is aligned with the center of the second subframe and is a mid-frame analysis window. If the encoder switches to transform coding, it is nevertheless preferable to transmit only one set of LPC coefficient data obtained from the LPC analysis based on the end-of-frame LPC analysis window. In addition, on the decoder side, it is preferable not to use this LPC data directly for the synthesis of transform coding, and in particular spectral weighting of TCX coefficients. Instead, it is preferable to interpolate the TCX data obtained from the LPC analysis window of the end of the frame of the current frame using the data obtained by the LPC analysis window of the end of the frame from the last frame, i.e., the frame immediately preceding the current frame in time. By transmitting only one set of LPC coefficients for an entire frame in TCX mode, an additional reduction in bit rate can be obtained compared to transmitting two sets of LPC coefficient data for mid-frame analysis and end frame analysis. However, when the encoder switches to ACELP mode, both sets of LPC coefficients are transmitted from the encoder to the decoder.

In addition, it is preferable that the mid-frame LPC analysis window terminates immediately at the next frame boundary of the current frame and extends further to the last frame. This does not introduce any delay since the last frame is already available and can be used without any delay.

On the other hand, it is preferable that the end-of-frame analysis window starts somewhere within the current frame, rather than at the beginning of the current frame. This, however, is not a problem, since the average of the end-of-frame LPC dataset for the last frame and the end-frame LPC dataset for the current frame are used to generate TCX weighting, so that in the end all the data is used in some sense to calculate the LPC coefficients. Therefore, the start of the end-of-frame analysis window is preferably within a portion of the leading view of the end-of-frame analysis window.

On the decoder side, a significantly reduced overhead is obtained for switching from one mode to another mode. The reason is that the non-overlapping part of the synthesis window, which is preferably symmetrical in its limits, is not associated with discrete values of the current frame, but is associated with discrete values of the future frame, and therefore extends only within the part of the leading view, that is, only in the future frame . Therefore, the synthesis window is such that only the first overlapping part, preferably starting in the middle part of the current frame, is within the current frame, and the second non-overlapping part extends from the end of the first overlapping part to the end of the current frame and, therefore, the second overlapping part coincides with the part leading view. Therefore, when there is a transition from TCX to ACELP, data obtained from the overlapping part of the synthesis window is simply discarded and replaced with predictive coding data, which are available from the very beginning of the future frame from the ACELP branch.

On the other hand, when there is a switch from ACELP to TCX, a special transition window is applied, which starts immediately at the beginning of the current frame, that is, the frame immediately after the switch, from the non-overlapping part, so that no data needs to be reconstructed to find overlapping "partners" . Instead, the non-overlapping portion of the synthesis window provides the correct data without any overlap and without any overlap-adding procedure required by the decoder. Only for overlapping parts, that is, the third part of the window for the current frame and the first part of the window for the next frame, the overlap-adding procedure is useful and is performed to have, as in direct MDCT, continuous increase / decrease from one block to another, in order to finally get good sound quality without the need to increase bit rate due to the critically sampled properties of MDCT, which is also known in the art under the term “time domain distortion correction (TDAC).

In addition, the decoder is useful in that for the ACELP encoding mode, LPC data received from the middle of the frame window and the end of the frame window in the encoder is transmitted, while for the TCX encoding mode, only one LPC data set obtained from the end of the frame window is used. However, for decoded spectrally weighted TCX data, the transmitted LPC data is not used as such, and the data is averaged with the corresponding data from the LPC analysis window of the end of the frame obtained for the last frame.

Preferred embodiments of the present invention are described below with reference to the accompanying drawings, in which:

FIG. 1a is a block diagram of a switchable audio encoder;

FIG. 1b is a block diagram of a corresponding switchable decoder;

FIG. 1c shows in more detail the transform parameter decoder shown in FIG. 1b;

FIG. 1d depicts in more detail the encoding mode with transform of the decoder in FIG. 1a;

FIG. 2a shows a preferred embodiment for a window function application block used in an encoder for LPC analysis on the one hand and transform coding analysis on the other hand, it is a representation of the synthesis window used in the transform coding decoder in FIG. 1b;

FIG. 2b shows a sequence of aligned LPC analysis windows and TCX windows for a time span of more than two frames;

FIG. 2c shows a situation for transition from TCX to ACELP and a transition window for transition from ACELP to TCX;

FIG. 3a shows in more detail the encoder of FIG. 1a;

FIG. 3b shows a synthesis analysis procedure for selecting a coding mode for a frame;

FIG. 3c shows an additional embodiment for choosing between modes for each frame;

FIG. 4a shows the calculation and use of LPC data obtained by using two different LPC analysis windows for the current frame;

FIG. 4b shows the use of LPC data obtained by processing using a window function using the LPC analysis window for a TCX encoder branch;

FIG. 5a shows an LPC analysis window for AMR-WB;

FIG. 5b shows symmetric windows for AMR-WB + for LPC analysis;

FIG. 5c shows LPC analysis windows for a G.718 encoder;

FIG. 5d shows the LPC analysis windows used in USAC; and

FIG. 6 shows a TCX window for the current frame relative to the LPC analysis window for the current frame.

FIG. 1a shows an apparatus for encoding an audio signal having a stream of discrete values of an audio signal. The arrival of the discrete values of the audio signal or audio data to the encoder is shown at number 100. The audio data is supplied to the window function application unit 102 for applying the predictive encoding analysis window to the stream of discrete audio values to obtain predicted analysis data processed by the window function. The window function application unit 102 is further configured to apply a transform coding analysis window to a stream of discrete audio signal values to obtain data processed by the window function for transform analysis. Depending on the embodiment, the LPC window is not applied directly to the original signal, but to the “pre-emphasized” signal (as in AMR-WB, AMR-WB +, G718 and USAC). On the other hand, the TCX window is applied directly to the original signal (as in USAC). However, both windows can also be applied to the same signals, or the TCX window can also be applied to the processed audio signal obtained from the original signal by pre-emphasis or any other weighting used to improve the quality or compression efficiency.

The conversion coding analysis window is associated with discrete values of the audio signal in the current frame of discrete values of the audio signal and with discrete values of the audio signal of a given part of the future frame of discrete values of the audio signal, which is part of the look-ahead coding with conversion.

In addition, the prediction encoding analysis window is associated with at least a portion of the discrete audio values of the current frame and with discrete audio values of a predetermined portion of the future frame that is part of the predictive encoding look-ahead.

As indicated in block 102, the conversion coding look-ahead part and the predictive coding look-ahead part are aligned with each other, which means that these parts are either identical or close enough to each other, for example, less than 20% differ from each other parts of the look-ahead coding review with prediction or less than 20% of the parts of the look-ahead coding with conversion. Preferably, when the leading parts of the look-ahead are identical or different from each other, less than 5% of the parts of the leading-look coding with predictions or less than 5% of the parts of the leading-look coding with conversion.

The encoder further comprises an encoding processor 104 for generating predicted encoded data for the current frame using windowed data for prediction analysis or for generating transform encoded data for the current frame using windowed data for analysis with conversion.

In addition, the encoder preferably comprises an output interface 106 for receiving for the current frame and, in fact, for each frame, LPC data and transform encoded data (such as TCX data) or predictive encoded data (ACELP data) on the line 108b. The coding processor 104 provides these two kinds of data and receives, as input, the window-processed data for the prediction analysis designated as 110a and the window-processed data for the conversion analysis designated 110b. In addition, the encoding device comprises an encoding mode selector or controller 112 that receives audio data 100 as input and which provides control information for encoding processor 104 via control lines 114a or control information for output interface 106 via control line 114b as input. .

FIG. 3a shows in more detail the encoding processor 104 and the window function application unit 102. The window function application unit 102 preferably comprises, as a first module, a predictive coding analysis window function or LPC application unit 102a and, as a second component or module, a transform coding window function application block 102b (such as a TCX window function application unit). As shown by arrow 300, the LPC analysis window and the TCX window are aligned with each other, so that the pieces of the look-ahead of both windows are identical to each other, which means that both parts of the look-ahead extend to the same point in time in a future frame. The upper branch in FIG. 3a from the LPC window function application block 102a and further to the right is a prediction encoding branch comprising an LPC analyzer and an interpolator 302, a perceptual weighting filter or a weighting unit 304 and a predictive encoding parameter calculator 306, such as an ACELP parameter calculator. Audio data 100 is provided for the LPC window function application unit 102a and the perceptual weighting unit 304. In addition, audio data is provided for the TCX window function application unit, and the lower branch from the output of the TCX window function application unit to the right constitutes a transform encoding branch. This transform coding branch comprises a time-frequency transform unit 310, a spectral weighting unit 312, and a processing / quantization encoding unit 314. The time-frequency transform block 310 is preferably implemented as a distortion transform, such as MDCT, MDST, or any other transform that has a number of input values greater than the number of output values. The time-frequency conversion has, as input, the window data processed by the window function using the TCX window function application unit or, generally speaking, transform coding.

Although FIG. 3a shows for the predictive coding branch LPC processing with the ACELP coding algorithm, other predictive encoders such as CELP or any other time-domain encoders known in the art may also be used, although the ACELP algorithm is preferred because of its quality on the one hand and its efficiency on the other hand.

Furthermore, for the transform coding branch in the time-frequency transform block 310, in particular, MDCT processing is preferable, although any other transforms in the spectral domain can also be performed.

In addition, FIG. 3a shows spectral weighting 312 for converting spectral values output by block 310 into an LPC region. This spectral weighting 312 is performed using the weighting data obtained from the LPC analysis data generated by block 302 in the predictive coding branch. Alternatively, however, the conversion from the time domain to the LPC area may also be performed in the time domain. In this case, the LPC analysis filter will be placed in front of the TCX window function application block 102b to calculate the residual time domain prediction data. However, it was found that the conversion from the time domain to the LPC region is preferably performed in the spectral region by spectrally weighting the transform encoded data using LPC analysis data converted from the LPC data to corresponding weights in the spectral region, such as the MDCT region.

FIG. 3b is a perspective view for illustrating analysis by synthesizing or determining a “feedback” coding mode for each frame. To this end, the encoder shown in FIG. 3c, comprises a complete transform coding encoder and a transform encoding decoder, as shown under 104b, and further comprises a complete predictive encoding encoder and a corresponding decoder, designated 104a in FIG. 3c. Both blocks 104a, 104b receive audio data as input and perform a complete encoding / decoding operation. Then, the results of the encoding / decoding operation for both encoding branches 104a, 104b are compared with the original signal and a quality indicator is determined to find out which encoding mode provides the best quality. The quality score can be a segmented SNR value or an average segmented SNR, such as, for example, described in section 5.2.3 of 3GPP TS 26.290. However, any other quality indicators that are usually based on comparing the encoding / decoding result with the original signal may also be applied.

Based on the quality score that is provided from each branch 104a, 104b to the decision block 112, the decision block decides whether the frame in question should be encoded using ACELP or TCX. After making a decision, there are several ways to make the choice of encoding mode. One way is that the decision block 112 controls the corresponding encoder / decoder blocks 104a, 104b to simply output the encoding result for the current frame to the output interface 106 so that it is guaranteed that for only one frame, only one encoding result is transmitted in the output encoded signal at 107.

Alternatively, both devices 104a, 104b may transmit their encoding result to the output interface 106, and both results are stored in the output interface 106 until the decision unit orders the output interface to either output the block 104b or block 104a through line 105.

FIG. 3b shows in more detail the concept of FIG. 3c. In particular, block 104a comprises a complete ACELP encoder, a complete ACELP decoder, and a comparison device 112a. The comparison device 112a provides a quality score for the comparison device 112c. The same is true for comparator 112b, which has a quality metric as a result of comparing the TCX encoded and decoded signal with the original audio signal. Then, both comparison devices 112a, 112b provide their quality metrics to the final comparison device 112c. Depending on which quality score is better, the comparison device selects CELP or TCX. The decision can be clarified by introducing additional factors into the decision making process.

Alternatively, a non-feedback mode may be performed to determine the encoding mode for the current frame based on the analysis of the audio signal for the current frame. In this case, the decision block 112 in FIG. 3c will perform an analysis of the audio data signal for the current frame and will then either control the ACELP encoder or the TCX encoder to actually encode the current audio frame. In this situation, the encoder does not need a full decoder, and an embodiment of only the encoding steps of the encoder will be sufficient. Signal classifications and decision making regarding feedback signals are also described, for example, in AMR-WB + (3 GPP TS 26.290).

FIG. 2a shows a preferred embodiment of a window function application unit 102 and, in particular, windows provided by a window function application unit.

Preferably, the predictive coding analysis window for the current frame is centered in the center of the fourth subframe, and this window is indicated by 200. In addition, it is preferable to use an additional LPC analysis window, that is, the mid-frame LPC analysis window, labeled 202 and having a center in the center second subframe of the current frame. In addition, a transform coding window, such as, for example, MDCT window 204, is positioned relative to two LPC analysis windows 200, 202, as shown in the figure. In particular, the look-ahead portion 206 from the analysis window has the same time length as the look-ahead portion 208 of the predictive coding analysis window. Both pieces of look-ahead extend for 10 ms in a future frame. In addition, it is preferable that the transform coding analysis window not only has an overlapping part 206, but also has an non-overlapping part 208 between 10 and 20 ms and a first overlapping part 210. The overlapping parts 206 and 210 are such that the overlapping adding unit in the decoder processes with adding overlap in the overlapping part, and the procedure for adding overlap is not necessary for the non-overlapping part.

Preferably, the first overlapping portion 210 starts at the beginning of the frame, i.e. at 0 ms, and extends to the center of the frame, i.e. 10 ms. In addition, the non-overlapping part extends from the end of the first part of the frame 210 to the end of the frame at 20 ms, so that the second overlapping part 206 completely coincides with the leading viewing part. This has the advantage of switching from one mode to another mode. In terms of TCX efficiency, it would be better to use a sine window with full overlap (20 ms overlap, as in USAC). This, however, would necessitate the use of technology, such as the direct correction of distortions for transitions between TCX and ACELP. USAC Direct Correction is used to eliminate distortion introduced by the absence of the following TCX frames (replaced by ACELP). Direct elimination of distortion requires a significant number of bits and, therefore, is not suitable for a constant bit rate and, in particular, a codec with a low bit rate, as a preferred embodiment of the described codec. Therefore, in accordance with embodiments of the invention, instead of using FAC (direct distortion elimination), the TCX window overlap is reduced and the window is shifted to the future so that the entire overlapping portion 206 is placed in a future frame. In addition, the window shown in FIG. 2a for transform coding, however, has a maximum overlap to obtain perfect reconstruction in the current frame when the next frame is ACELP, without using direct distortion correction. This maximum overlap is preferably set to 10 ms, which is an available leading-time view, i.e. 10 ms, as can be seen from FIG. 2a.

Although FIG. 2a has been described with reference to an encoder in which the transform encoding window 204 is an analysis window, it should be noted that the window 204 is also a synthesis window for transform decoding. In a preferred embodiment, the analysis window is identical to the synthesis window, and both windows are symmetrical on their own. This means that both windows are symmetrical about the (horizontal) center line. In other applications, however, asymmetric windows may be used where the analysis window differs in shape from the synthesis window.

FIG. 2b depicts a sequence of windows on a portion of a past frame, subsequent to the current frame, future frame that follows the current frame, and next to the future frame that follows the future frame.

It becomes clear that the overlap-adding part, processed by the overlap-adding processor, depicted at number 250, extends from the beginning of each frame to the middle of each frame, that is, between 20 and 30 ms for calculating future frame data and between 40 and 50 ms for calculating TCX data for a frame following the future, or between zero and 10 ms to calculate data for the current frame. However, to calculate the data in the second half of each frame, there is no added overlap, and therefore no technique for directly eliminating distortion is needed. This is due to the fact that the synthesis window has a non-overlapping part in the second half of each frame.

Typically, the MDCT window length is twice the frame length. This also occurs in the present invention. However, when, again, FIG. 2a, it becomes clear that the analysis / synthesis window extends only from zero to 30 ms, and the total window length is 40 ms. This full length is of great importance in providing input for the corresponding convolution and inverse convolution of the MDCT calculation. To expand the window to a full length of 40 ms, 5 ms of zero values are added between -5 and 0 ms, and 5 ms of zero MDCT values are also added at the end of the frame between 30 and 35 ms. However, these additional parts, with only zeros, do not play any role when it comes to considering the delay, since the encoder or decoder knows that the last five ms of the window and the first five ms of the window are zeros, so this data is already present without any or delays.

FIG. 2c shows two possible transitions. For the transition from TCX to ACELP, however, no special precautions are necessary, since when assumed with respect to FIG. 2a that the future frame is an ACELP frame, the data obtained by TCX decoding the last frame for the look-ahead part 206 can simply be deleted, since the ACELP frame starts immediately at the beginning of the future frame and, therefore, there is no gap without data. ACELP data is self-consistent and, therefore, the decoder, when switching from TCX to ACELP, uses the data calculated by TCX for the current frame, discards the data obtained by TCX processing for the future frame, and instead uses the data of the future frame from the ACELP branch.

However, when the transition from ACELP to TCX is performed, a special transition window is used, as shown in FIG. 2c. This window starts at the beginning of the frame from zero to 1, has a non-overlapping portion 220 and has an overlapping portion at the end, designated 222, which is identical to the overlapping portion 206 of the direct MDCT window.

This window is additionally filled with zeros between -12.5 ms and zero at the beginning of the window and between 30 and 35.5 ms at the end, that is, after part 222 of the forward viewing. This results in an increased conversion length. The length is 50 ms, but the length of the direct analysis / synthesis window is only 40 ms. This, however, does not decrease efficiency or increase bit rate, and this longer conversion is necessary when switching from ACELP to TCX takes place. The transition window used in the corresponding decoder is identical to the window shown in FIG. 2c.

The decoder is discussed in more detail below. FIG. 1b shows an audio decoder for decoding an encoded audio signal. The audio decoder comprises a prediction parameter decoder 180, wherein the prediction parameter decoder is configured to decode data for the prediction encoded frame from the encoded audio signal received at 181 and supplied to the interface 182. The decoder further comprises a transform parameter decoder 183 to decode data for the frame encoded with conversion, the audio signal on line 181. The conversion parameter decoder is configured to perform, preferably subject to distortions of the spectral-temporal transformation and for applying the synthesis window to the converted data to obtain data for the current frame and future frame. The synthesis window has a first overlapping portion, an adjacent second non-overlapping portion, and an adjacent third overlapping portion, as shown in FIG. 2a, while the third overlapping part is associated only with discrete values of the audio signal for the future frame, and the non-overlapping part is associated only with the data of the current frame. In addition, an overlapping block 184 is provided for superimposing and adding discrete synthesis window values associated with a third overlapping part of the synthesis window for the current frame and a synthesis window with discrete values associated with the first overlapping part of the synthesis window for a future frame to obtain the first discrete part audio values for the future frame. The remaining discrete values of the audio signal for the future frame are the discrete values processed using the synthesis window function associated with the second non-overlapping part of the synthesis window for the future frame obtained without adding overlap when the current frame and the future frame contain data encoded with conversion. However, when switching from one frame to the next frame takes place, combiner 185 is useful, which should take care of a good switch from one encoding mode to another encoding mode, in order to finally receive decoded audio data at the output of combiner 185.

FIG. 1c shows in more detail the construction of a transform parameter decoder 183.

The decoder comprises a decoder processing stage 183a that is configured to perform all the processing necessary to decode the encoded spectral data, such as arithmetic decoding, or Huffman decoding, or, in general, statistical decoding and subsequent inverse quantization, noise filling, etc. d. to obtain decoded spectral values at the output of block 183. These spectral values are supplied to the input of spectral weighting unit 183b. The spectral weighting unit 183b receives spectral weighting data from the LPC weighted data calculator 183c, to which LPC data generated by the prediction analysis unit on the encoder side and received at the decoder via the input interface 182 is supplied. The inverse spectral conversion is then performed, which contains, as a first step, preferably, the inverse transform 183d of DCT-IV and the subsequent inverse convolution and processing 183e using the window synthesis function, before the data for the future frame, for example, redostavleny added overlap unit 184. Adding unit may perform an overlap add operation overlap when data is available for the next future frame. Blocks 183d and 183e together constitute spectral / temporal conversion, or, in the embodiment of FIG. 1c, the preferred inverse transformation is MDCT (MDCT ^-1 ).

In particular, block 183d receives data for a 20 ms frame and increases the amount of data in the reverse convolution step of block 183e to data for 40 ms, that is, twice as much data as before, and then to that 40 ms of data, a synthesis window is applied having a duration of 40 ms (when zero parts are added at the beginning and end of the window). Then, at the output of block 183e, data is available for the current block and data within the look-ahead part for the future block.

FIG. 1d shows the corresponding processing on the encoder side. The functions discussed in the context of FIG. 1d are implemented in coding processor 104 or corresponding blocks in FIG. 3a. Time-frequency conversion 310 of FIG. 3a is preferably implemented as an MDCT and comprises a step 310a of applying a window function, a convolution, where the operation of applying a window function in block 310a is implemented as a TCX window function application block 103d. Therefore, in fact, the first operation in block 310 in FIG. 3a is a convolution operation to return 40 ms of input data to 20 ms frame data. Then, with the collapsed data that has now received the distortion contribution, DCT-IV is executed, as shown in block 310d. Block 302 (LPC analysis) provides LPC data obtained by analysis using the end-of-frame LPC window to block 302b (LPC to MDCT), and block 302d generates weights for performing spectral weighting using spectral weighting unit 312. Preferably, 16 LPCs for one 20 ms frame in TCX coding mode are converted to 16 weighting factors of the MDCT domain, preferably by using oDFT (odd discrete Fourier transform). For other modes, such as NB modes having a sampling frequency of 8 kHz, the number of LPC coefficients may be lower, for example, 10. For other modes with higher sampling rates, there may also be more than 16 LPC coefficients. The result of this oDFT is 16 weight values, and each weight value is associated with a band of spectral data obtained by block 310b. Spectral weighting is done by dividing all the MDCT spectral values for one band by the same weight value associated with that band to perform this spectral weighting operation very efficiently in block 312. Therefore, each of the 16 bands of MDCT values is divided by a corresponding weight coefficient so that derive spectrally weighted spectral values, which are then further processed by block 314, as is known in the art, i.e., for example, by quantization and statistical coding Ania.

On the decoder side, on the other hand, spectral weighting corresponding to block 312 in FIG. 1d will be the multiplication performed by the spectral weighting unit 183b shown in FIG. 1s

Next, FIG. 4a and FIG. 4b to show how the LPC data generated by the LPC analysis window or generated by the two LPC analysis windows shown in FIG. 2, are used in ACELP mode or in TCX / MDCT mode.

After applying the LPC analysis window, autocorrelation is calculated using the LPC data processed using the window function. Then the Levinson Darbin algorithm is applied to the autocorrelation function. Then, 16 LP coefficients for each LP analysis, that is, 16 coefficients for the mid-frame window and 16 coefficients for the end-frame window, are converted to ISP values. Therefore, the steps from calculating autocorrelation to ISP conversion are performed, for example, in block 400 of FIG. 4a.

Then the calculations are continued on the encoder side in the form of quantization of the ISP coefficients. Then, the ISP coefficients are again inverse quantized and converted back to the LP coefficient domain. Therefore, LPC data or, in other words, 16 LPC coefficients is obtained, which are slightly different from the LPC coefficients obtained in block 400 (due to quantization and inverse quantization), and which can then be used directly for the fourth subframe, as shown in step 401 For other subframes, however, it is preferable to perform several interpolations as, for example, described in section 6.8.3 of ITU-T Rec. ITU-T G.718 (06/2008). The LPC data for the third subframe is calculated by interpolating the LPC data of the end of the frame and the middle of the frame represented by block 402. The preferred interpolation is to divide all the corresponding data into two and add them, that is, the average of the LPC data of the end of the frame and the middle of the frame. In order to calculate the LPC data for the second subframe, as shown in block 403, interpolation is further performed. In particular, 10% of the end-of-frame LPC data values of the last frame, 80% of the mid-frame LPC data of the current frame and 10% of the LPC data values for the end of the current frame frame are used to ultimately calculate the LPC data for the second subframe.

Finally, LPC data for the first subframe is calculated, as shown in block 404, by forming the middle between the LPC data of the end frame of the last frame and the LPC data of the middle of the frame of the current frame.

To perform ACELP encoding, both quantized sets of LPC parameters, i.e., from mid-frame analysis and end-of-frame analysis, are transmitted to the decoder.

Based on the results for the individual subframes computed by blocks 401-404, ACELP calculations are performed, as indicated in block 405, to obtain ACELP data to be transmitted to the decoder.

Next, FIG. 4b. Again, in block 400, LPC data of the middle of the frame and the end of the frame are calculated. However, since we have the TCX coding mode, only the end-of-frame LPC data is transmitted to the decoder, and the mid-frame LPC data is not transmitted to the decoder. In particular, the LPC coefficients are not directly transmitted to the decoder, but the values obtained after ISP conversion and quantization are transmitted. Therefore, it is preferable that, like the LPC data, the quantized ISP values obtained from the LPC data coefficients of the end of the frame are transmitted to the decoder.

However, the procedures of steps 406-408 must be performed in the encoder to obtain weights to obtain weights for weighting the MDCT spectral data of the current frame. To this end, the LPC data of the end of the frame of the current frame and the LPC data of the end of the frame of the previous frame are interpolated. However, it is preferable not to directly interpolate the LPC data coefficients obtained directly from the LPC analysis. Instead, it is preferable to interpolate the quantized and again quantized ISP values obtained from the corresponding LPC coefficients. Therefore, the LPC data used in block 406, as well as the LPC data used for other calculations in block 401-404, is always preferably the quantized and inverse quantized ISP data obtained from the original 16 LPC coefficients per analysis window LPC

The interpolation in block 406 is preferably a pure averaging, i.e. the corresponding values are added and divided by two. Then, in block 407, the MDCT spectral data of the current frame is weighted using the interpolated LPC data and, in block 408, further processing of the weighted spectral data is performed to finally obtain encoded spectral data that will be transmitted from the encoder to the decoder. Therefore, the procedures performed at step 407 correspond to block 312, and the procedure performed at block 408 in FIG. 4b corresponds to block 314 in FIG. 3a. The corresponding operations are actually performed on the side of the decoder. Therefore, the same interpolations are needed on the decoder side in order to calculate the spectral weighting coefficients on the one hand or to calculate the LPC coefficients for individual subframes by interpolation on the other hand. Therefore, FIG. 4a and FIG. 4b are equally applicable to the decoder side with respect to the procedures in blocks 401-404 or 406 of FIG. 4b.

The present invention is particularly useful for low latency codec embodiments. This means that such codecs are designed to have an algorithmic or systematic delay of preferably less than 45 ms and, in some cases, even equal to or less than 35 ms. However, part of the lead-in viewing for LPC analysis and TCX analysis is necessary to get good sound quality. Therefore, a good compromise is needed between these two conflicting requirements. It was found that a good compromise between delay on the one hand and quality on the other can be obtained by a switchable audio encoder or decoder having a frame duration of 20 ms, but it was also found that values for frame lengths between 15 and 30 ms also provide acceptable results. . On the other hand, it was found that the 10 ms look-ahead portion is acceptable when it comes to delay problems, but values between 5 ms and 20 ms are also useful depending on the respective application. In addition, it was found that the ratio between the portion of the look-ahead and the frame length is useful when it is 0.5, but other values between 0.4 and 0.6 are also useful. Furthermore, although the invention has been described with ACELP on the one hand and MDCT-TCX on the other, other time-domain algorithms such as CELP or any other prediction or waveform algorithms are also useful. For TCX / MDCT, other transform domain coding algorithms, such as MDST or any other transform-based algorithms, may also be used.

The same is true for a particular embodiment of LPC analysis and LPC calculation. It is preferable to rely on the procedures described above, but other procedures for calculation / interpolation and analysis can also be used if these procedures rely on the LPC analysis window.

Although some aspects have been described in the context of the device, it is obvious that these aspects also represent a description of the corresponding method, where the unit or device corresponds to a method step or a property of a method step. Similarly, the aspects described in the context of a method step also provide a description of the corresponding unit or element or property of the corresponding device.

Depending on some requirements of an embodiment, embodiments of the invention may be implemented in hardware or in software. An embodiment may be performed using a digital storage medium such as a floppy disk, DVD, CD, ROM, PROM, EPROM, EEPROM or flash memory having electronically readable control signals that interact (or are capable of interacting) with the programmable a computer system such that an appropriate method is performed.

Some embodiments of the present invention comprise a non-transitory storage medium having electronically readable control signals that can interact with a programmable computer system such that one of the methods described herein is performed.

Typically, embodiments of the present invention may be implemented as a computer program product with program code, program code executed to execute one of the methods when the computer program product is executed on a computer. The program code may, for example, be stored on a computer-readable medium.

Other embodiments include a computer program for performing one of the methods described herein, which is stored on a computer-readable medium.

Therefore, in other words, an embodiment of the method according to the invention is a computer program having program code for executing one of the methods described herein when the computer program is executed on a computer.

Another embodiment of the methods of the invention is a storage medium (either a digital storage medium or a computer-readable medium) comprising a computer program recorded thereon for performing one of the methods described herein.

Another embodiment of the method according to the invention is a data stream or a sequence of signals representing a computer program for performing one of the methods described herein. The data stream or signal sequence may, for example, be configured to be transmitted via a data connection, for example via the Internet.

Another embodiment includes a processing means, such as a computer or programmable logic device, configured or adapted to perform one of the methods described herein.

Another embodiment includes a computer having a computer program installed thereon for performing one of the methods described herein.

In some embodiments, a programmable logic device (eg, a user programmable gate array) may be used to perform some or all of the functions of the methods described herein. In some embodiments, a user-programmable gate array may interact with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by any hardware.

The embodiments described above merely illustrate the principles of the present invention. It is understood that modifications and variations of the arrangements and details described herein will be apparent to those skilled in the art. Therefore, it is intended that the invention be limited only by the scope of the appended claims, and not by the specific details presented herein in the form of a description and explanation of embodiments.

Claims (26)

1. A device for encoding an audio signal having a stream of discrete values (100) of an audio signal, comprising:
a window function application unit (102) for applying a prediction coding analysis window (200) to a stream of discrete audio signal values to obtain data processed by the window function for prediction analysis and for applying a time-frequency transformation coding window (204) to the stream of discrete values of the audio signal to obtain processed data using the window function for analysis with conversion, wherein the prediction encoding is the prediction encoding of the future frame,
wherein the conversion coding analysis window is associated with discrete values of the audio signal within the current frame of discrete values of the audio signal and with discrete values of the audio signal of a predetermined portion of the future frame of discrete values of the audio signal, which are part (206) of the forward looking encoding with conversion,
moreover, the prediction coding analysis window is associated with at least a portion of the discrete audio values of the current frame and with discrete audio values of a predetermined portion of the future frame, which are part (208) of the predictive coding preview,
moreover, the part (206) of the forward lookup of the coding with conversion and the part (208) of the forward lookup of the coding with prediction are identical to each other or differ from each other by less than 20% of the part (208) of the forward lookup of the coding with prediction or less than 20% of the part ( 206) look-ahead coding conversion; and
an encoding processor (104) for generating predicted encoded data for the current frame using windowed data for prediction analysis or for generating transform encoded data for the current frame using windowed data for analysis with conversion. 2. The device according to claim 1, wherein the transform coding analysis window (204) comprises a non-overlapping portion extending in a look-ahead coding conversion portion (206). 3. The device according to claim 1, wherein the transform coding analysis window (204) comprises an additional overlapping part (210) starting at the beginning of the current frame and ending at the beginning of the non-overlapping part (208). 4. The device according to claim 1, in which the block (102) for applying the window function is configured to use only the initial window (220, 222) to switch from predictive encoding to encoding with conversion from a frame to the next frame, and the initial window is not used to move from transform coding to predictive coding from one frame to the next frame. 5. The device according to claim 1, further comprising:
an output interface (106) for outputting an encoded signal for the current frame; and
an encoding mode selector (112) for controlling an encoding processor (104) for outputting either prediction encoded data or transform encoded data for the current frame,
wherein the encoding mode selector (112) is configured to switch only between prediction encoding or transform coding for the whole frame, so that the encoded signal for the whole frame either contains prediction encoded data or transform encoded data. 6. The device according to claim 1,
in which the application block (102) of the window function uses, in addition to the prediction encoding analysis window, an additional prediction encoding analysis window (202) associated with discrete audio values located at the beginning of the current frame, and in which the encoding analysis window (200) with prediction is not associated with discrete values of the audio signal located at the beginning of the current frame. 7. The device according to claim 1,
wherein the frame contains a plurality of subframes, the prediction analysis window (200) being centered in the center of the subframe, and the transform coding analysis window being centered on the boundary between the two subframes. 8. The device according to p. 7,
wherein the prediction analysis window (200) is centered in the center of the last subframe of the frame, wherein the additional analysis window (202) is centered in the center of the second subframe of the current frame, and wherein the transform coding analysis window is centered on the boundary between the third and fourth subframe of the current frame, the current frame is subdivided into four subframes. 9. The device according to claim 1, in which the additional window (202) of the prediction coding analysis does not have a part of the look-ahead in the future frame and is associated with discrete values of the current frame. 10. The device according to claim 1, wherein the transform coding analysis window further comprises a zero part before the start of the window and a zero part after the end of the window so that the total time length of the transform coding analysis window is two times the length in time of the current frame . 11. The device according to claim 10, in which the transition window is used to switch from the prediction encoding mode to the encoding mode with conversion from one frame to the next frame by the window function application unit (102),
moreover, the transition window contains the first non-overlapping part, starting at the beginning of the frame, and an overlapping part, starting at the end of the non-overlapping part and extending to the future frame,
moreover, the overlapping part, extending into the future frame, has a length that is identical to the length of the leading part of the coding preview with the transformation of the analysis window. 12. The device according to claim 1, wherein the time length of the transform coding analysis window is greater than the time length of the predictive coding analysis window (200, 202). 13. The device according to claim 1, further comprising:
an output interface (106) for outputting an encoded signal for the current frame; and
an encoding mode selector (112) for controlling an encoding processor (104) for outputting either prediction encoded data or transform encoded data for the current frame,
moreover, the window (102) is configured to use an additional prediction encoding window located in the current frame in front of the prediction encoding window, and
wherein the encoding mode selector (112) is configured to control the encoding processor (104) to send only predictive encoding analysis data obtained from the predictive encoding window when the encoded transform data is output to the output interface and not send encoding analysis data with prediction obtained from the additional prediction coding window, and
wherein the encoding mode selector (112) is configured to control the encoding processor (104) to send prediction encoding analysis data obtained from the prediction encoding window and to send prediction encoding analysis data obtained from the additional prediction encoding window when the data encoded with prediction are output to the output interface. 14. The device according to claim 1, wherein the encoding processor (104) comprises:
a prediction encoding analyzer (302) for obtaining prediction encoding data for a current frame from the window-processed data (100a) for prediction analysis;
a predictive coding branch comprising:
a filter stage (304) for computing filter data from discrete audio signal values for the current frame using predictive coding data; and
a predictor encoder parameter calculator (306) for calculating predictive encoding parameters for the current frames; and
a transform encoding branch containing:
spectral-time converter (310) for
transforming window data for a coding algorithm with conversion to a spectral representation;
a spectral weighting unit (312) for weighting the spectral data using the weighted weighting data obtained from the predictive coding data to obtain weighted spectral data; and
a spectral data processor (314) for processing weighted spectral data to obtain transform encoded data for the current frame. 15. A method of encoding an audio signal having a stream of discrete values (100) of an audio signal, comprising:
application (102) of a predictive coding analysis window (200) to a stream of discrete values of an audio signal to obtain processed data using a window function for predicting analysis and applying a time-frequency coding analysis window (204) to a stream of discrete values of an audio signal for receiving processed by a window function data for analysis with conversion, while prediction encoding is prediction encoding of a future frame,
wherein the conversion coding analysis window is associated with discrete values of the audio signal within the current frame of discrete values of the audio signal and with discrete values of the audio signal of a predetermined portion of the future frame of discrete values of the audio signal, which are part (206) of the forward looking encoding with conversion,
moreover, the prediction coding analysis window is associated with at least a portion of the discrete audio values of the current frame and with discrete audio values of a predetermined portion of the future frame, which are part (208) of the predictive coding preview,
moreover, the part (206) of the forward lookup of the coding with conversion and the part (208) of the forward lookup of the coding with prediction are identical to each other or differ from each other by less than 20% of the part (208) of the forward lookup of the coding with prediction or less than 20% of the part ( 206) look-ahead coding conversion; and
generating (104) transform encoded data for the current frame using the window-processed data for prediction analysis, or to generate transform encoded data for the current frame using the window-processed data for transform . 16. An audio decoder for decoding an encoded audio signal, comprising:
a prediction parameter decoder (180) for decoding data for a prediction encoded frame from the encoded audio signal, wherein the prediction encoded frame is a prediction encoded frame;
a transform parameter decoder (183) for decoding data for a frame encoded with a time-frequency transform from the encoded audio signal,
moreover, the transform parameter decoder (183) is configured to perform spectral-temporal conversion and apply the synthesis window to the converted data to obtain data for the current frame and the future frame, the synthesis window having a first overlapping part, an adjacent second non-overlapping part and an adjacent third overlapping part (206) moreover, the third overlapping part is associated with discrete values of the audio signal for the future frame, and the non-overlapping part (208) is associated with the data of the current frame; and
block adding (184) for overlapping and adding discrete values processed using the synthesis window associated with the third overlapping part of the synthesis window for the current frame and processed with the synthesis window discrete values associated with the first overlapping part of the synthesis window for the future frame, receive the first part of the discrete values of the audio signal for the future frame, and the remaining discrete values of the audio signal for the future frame are processed using the synthesis window discrete values values associated with the second non-overlapping part of the synthesis window for the future frame, obtained without adding overlap, when the current frame and the future frame contain data encoded with the conversion. 17. The audio decoder according to claim 16, in which the current frame of the encoded audio signal contains data encoded with conversion, and the future frame contains data encoded with prediction, and the decoder (183) of the conversion parameters is configured to perform processing using the synthesis window using the synthesis window for the current frame, in order to obtain discrete values of the audio signal processed using the window function associated with the non-overlapping part (208) of the synthesis window, and discrete values processed using the synthesis window an audio signal associated with the third part of the overlapped synthesis window to the current frame is discarded, and wherein discrete values for the future audio signal frames are provided by the decoder (180) prediction parameters without data from the decoder (183) transformation parameters. 18. The audio decoder according to claim 16, in which the current frame contains predictive encoding data, and the future frame contains transform encoding data,
in which the transform parameter decoder (183) is configured to use a transition window different from the synthesis window, in which the transition window (220, 222) contains the first non-overlapping part (220) at the beginning of the future frame and the overlapping part (222) starting at the end of the future frame and extending into the frame following the future frame in time, and
in which discrete values of the audio signal for the future frame are generated without overlapping, and the audio data associated with the second overlapping window portion (222) for the future frame is calculated by adding overlapping block (184) using the first overlapping portion of the synthesis window for the frame following the future frame. 19. The audio decoder according to claim 16,
in which the transmitter (183) of the conversion parameters contains:
a spectral weighting unit (183b) for weighting the transform-decoded spectral data for the current frame using predictive encoding data; and
a predictive encoding weighting data calculator (183c) for calculating predictive encoding data by combining a weighted sum of the predicted encoding data obtained from the previous frame and predictive encoding data obtained from the current frame to obtain interpolated predictive encoding data. 20. The audio decoder according to claim 19,
in which the predictive encoding weighting data calculator (183c) is configured to convert the predictive encoding data into a spectral representation having weight values for each frequency range, and
wherein the spectral weighting unit (183b) is configured to weight all spectral values in a frequency band by means of an identical weight value for that frequency band. 21. The audio decoder according to any one of p. 16, in which the synthesis window is configured to have a total duration of less than 50 ms and more than 25 ms, the first and third overlapping parts have the same length, and the third overlapping part has a length of less than 15 ms. 22. The audio decoder according to claim 16,
in which the synthesis window has a duration of 30 ms without parts filled with zeros, the first and third overlapping parts have a duration of 10 ms each, and the non-overlapping part has a duration of 10 ms. 23. The audio decoder according to claim 16,
wherein the transform parameter decoder (183) is configured to apply a DCT transform (183d) having a number of discrete values corresponding to a frame length for a spectral-temporal transform, and a reverse convolution operation (183e) to generate a number of time values twice the number time values to DCT, and
apply (183e) the synthesis window to the result of the reverse convolution operation, and the synthesis window contains zero parts having a length equal to half the length of the first and third overlapping parts before the first overlapping part and after the third overlapping part. 24. A method for decoding an encoded audio signal, comprising:
performing (180) decoding the data for the prediction encoded frame from the encoded audio signal, wherein the prediction encoded frame is a prediction encoded future frame;
performing (183) decoding data for a frame encoded with a time-frequency transformation from the encoded audio signal,
moreover, the step of executing (183) decoding the data for the frame encoded with the conversion comprises performing a spectral-time conversion and applying the synthesis window to the converted data to obtain data for the current frame and the future frame, the synthesis window having a first overlapping part, an adjacent second non-overlapping part and an adjacent third overlapping part (206), while the third overlapping part is associated with discrete values of the audio signal for the future frame, and the non-overlapping part (208) is connected and a current frame data; and
overlapping and adding (184) the discrete values processed by the synthesis window associated with the third overlapping part of the synthesis window for the current frame and the processed by the synthesis window discrete values associated with the first overlapping part of the synthesis window for the future frame to obtain the first part of the discrete values of the audio signal for the future frame, and the remaining discrete values of the audio signal for the future frame are processed using the synthesis window, discrete values associated with the second the overlapping part of the synthesis window to the future frame, obtained without the addition of overlap, when the current frame and the next frame contains data encoded with the conversion. 25. A computer-readable medium containing computer-readable instructions stored on it that, when executed on a computer, force the computer to implement the audio encoding method of claim 15. 26. A computer-readable medium containing computer-readable instructions stored on it that, when executed on a computer, force the computer to implement a method for decoding an encoded audio signal according to claim 24.

Images