Interharmonic noise attenuation post-filter

Abstract

To provide a method and device for audio encoding and decoding of signals with mixed components from different sound sources.SOLUTION: A method includes: an upper switch 541, which is an interharmonic noise attenuation post-filter 500 that receives an input signal containing a preliminary audio signal decoded according to one of multiple decoding modes and supplies an output audio signal, that filters the preliminary audio signal to obtain a filtered signal by a post-filter 540, and selectively operates either a filtering mode that supplies the filtered signal as the output audio signal or a pass-through mode that supplies the unfiltered preliminary audio signal 544 as the output audio signal; and a control unit that receives the post-filtered signal and disables the post-filter while it is operating in pass-through mode, according to the post-filtered signal.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates generally to digital audio coding, and more particularly to a coding technique for audio signals containing components of different characters. [Background technology]

[0002] A widely used class of coding methods for audio signals, including speech and singing, includes CELP (code excited linear prediction), which uses multiple different coding methods, including frequency-domain coding methods specifically adapted for music, as well as methods of a more general nature, switching over time to accommodate changes in character between successive periods of the audio signal. For example, a simplified MPEG (Moving Pictures Experts Group) USAC (Unified Speech and Audio Coding; see ISO / IEC 23003-3 standard) decoder operates in at least three coding modes, namely AAC (Advanced Audio Coding; see ISO / IEC 13818-7 standard), ACELP (algebraic CELP), and TCX (transform-coded excitation), as shown in the top part of Figure 2.

[0003] Various implementations of CELP are adapted to the characteristics of the human speech system and, in some cases, to the human hearing. In this application, CELP refers to all possible implementations and variations, including, but not limited to, ACELP, wide- and narrow-band CELP, sub-band CELP (SB-CELP), low- and high-rate CELP, relaxed CELP (RCELP), low-delay CELP (LD-CELP), conjugate-structure CELP (CS-CELP), conjugate-structure ACELP (CS-ACELP), pitch-synchronous innovation CELP (PSI-CELP), and vector sum excited linear prediction (VSELP). The principles of CELP are explained in [Non-Patent Document 1], and some of its applications are described in References 25-29 cited in [Non-Patent Document 2]. As detailed in the former paper, a CELP decoder (or, equivalently, a CELP speech synthesizer) includes a pitch predictor and a pulse codebook. The pitch predictor recovers the periodic components of the coded speech signal. From the pulse codebook, an innovation sequence is added. The pitch predictor includes a long-delay predictor that recovers the pitch and a short-delay predictor that recovers the formants by spectral envelope shaping. Here, pitch is generally understood as the fundamental frequency of the tonal sound produced by the vocal cords and further colored by the resonant parts of the vocal tract. This frequency, together with its harmonics, characterizes speech and singing. Generally speaking, CELP methods are best suited to processing solo or single-part singing where the pitch frequency is well-defined and relatively easy to determine.

[0004] To improve the perceptual quality of CELP-coded speech, the CELP method is often combined with post-filtering (or, in other words, pitch enhancement). Patent document 1 and section 2 of non-patent document 2 disclose the desirable properties of such a post-filter, namely, the suppression of noise components between the harmonics of the detected voice pitch (long-term part; see section 4). It is believed that a significant part of this noise comes from the spectral envelope shaping. As shown in Figures 1 and 2, the long-term part of a simple post-filter has the transfer function

number

number

number

[0005] Only the long-term part described in the previous paragraph may be used, or it may be configured in series with a noise-shaping filter that preserves components in the frequency interval corresponding to the formants and attenuates noise in other spectral regions (short-term part; see section 3), i.e., in the "spectral valleys" of the formant envelope. In another possible variant, this filter set is further complemented by a progressive high-pass type filter that reduces the perceived degradation due to the spectral tilt of the short-term part.

[0006] Audio signals containing a mixture of components with different origins, such as tonal and non-tonal, vocal, instrumental, and non-musical components, cannot always be reproduced satisfactorily using existing digital coding techniques. More specifically, existing techniques are ill-equipped to handle such heterogeneous audio material, and may be suitable for some components but not others. In particular, music containing vocals and one or more instruments or choruses, coded using methods of the nature described above, often results in perceptible artifacts when decoded, which detract from the listening experience. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] U.S. Patent No. 4,969,192 [Non-patent literature]

[0008] [Non-Patent Document 1] R. Schroeder and S. Atal in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), vol. 10, pp. 937-940, 1985 [Non-patent document 2] Chen and Gersho, IEEE Transactions on Speech and Audio Processing, vol. 3, no. 1, 1995 [Non-patent document 3] Technical Specification ETSI TS 126 290, version 6.3.0, release 6 Summary of the Invention [Problem to be solved by the invention]

[0009] In order to alleviate at least some of the drawbacks described in the previous section, it is an object of the present invention to provide a method and an apparatus adapted for audio coding and decoding of signals containing a mixture of components from different sound sources. In particular, the present invention seeks to provide a method and an apparatus that is suitable in terms of coding efficiency or (perceived) reproduction fidelity or both. [Means for solving the problem]

[0010] The present invention achieves at least one of these objects by providing an encoder system, a decoder system, an encoding method, a decoding method and a computer program product for performing each of these methods, as defined in the independent claims. The dependent claims define embodiments of the invention.

[0011] The inventors have recognized that some artifacts perceived in coded audio signals of non-homogeneous origin are caused by inappropriate switching between multiple coding modes, at least one of which includes post-filtering in the decoder and at least one of which does not. More precisely, available post-filters remove not only harmonic noise (and, in some cases, spectral valley noise), but also signal components representing instruments, vocal accompaniment, and other material with "desired" characteristics. The fact that a measurable difference of 10 dB in spectral valleys (as described in Ghitza and Goldstein, IEEE Trans. Acoust, Speech, Signal Processing, vol. ASSP-4, pp. 697-708, 1986) is considered by many designers as justification for heavily filtering these frequency bands. However, the degradation in sound quality due to inter-harmonic (and spectral valley) attenuation itself is less significant than the degradation in sound quality that occurs when switching. Switching on the postfilter suddenly reduces the background of the singing voice, while deactivating the filter quickly increases the background. Frequent switching can result in switching artifacts, either due to the nature of the audio signal or the coding device's configuration. As an example, a USAC decoder can operate in ACELP mode combined with postfiltering, or in TCX mode without postfiltering. ACELP mode is used when there is a dominant vocal component. Thus, switching to ACELP mode is triggered at the beginning of a song, such as the beginning of a new musical phrase, the beginning of a new verse, or after the accompaniment drowns out the singing voice in a way that makes the vocal component less prominent. Experiments have confirmed that alternative solutions that use TCX coding throughout, or rather, bypassing the problem, do not solve the problem because they introduce reverberation-like artifacts.

[0012] Thus, in a first and second aspect, the present invention provides an audio coding method (and correspondingly functional audio coding system) characterized in that a device for decoding the bitstream output by the coding method decides whether to apply post-filtering including attenuation of inter-harmonic noise, the result of which is encoded in the bitstream and is accessible to the decoding device.

[0013] According to the present invention, the decision to use a post-filter is made separately from the decision regarding the most suitable coding mode, which allows the post-filtering state to be maintained long enough that switching does not bother the listener. Thus, the present coding method instructs the post-filter to remain inactive even when switching to a coding mode in which the filter was previously active.

[0014] It should be noted that the decision to apply post-filtering is typically made on a frame-by-frame basis. Thus, first, post-filtering is not applied to less than one frame at a time. Second, the decision to disable post-filtering is valid only for the current frame and may be maintained or re-evaluated for subsequent frames. In coding formats that allow for a main frame format and a short format that is a fraction of the normal format (e.g., 1 / 8 the length), post-filtering decisions do not need to be made for each short frame. Instead, the average or median of multiple short frames that together make up the normal frame may be calculated to determine the parameters for the filtering decision.

[0015] In third and fourth aspects of the present invention, there are provided audio decoding methods (and correspondingly functional audio decoding systems) comprising a decoding step followed by a post-filtering stage, including inter-harmonic noise attenuation, and comprising disabling the post-filter according to post-filtering information encoded in the bitstream signal.

[0016] A decoding method with these features is suitable for coding of mixed-source audio signals, since it has the ability to deactivate the postfilter solely based on the postfiltering information, and therefore independently of factors such as the current coding mode. When applied to coding methods where the postfilter function is traditionally attached to a specific coding mode, the ability to disable postfiltering allows a new mode of operation, namely, applying the traditional filtered decoding mode without filtering.

[0017] In yet another aspect, the present invention also provides a computer program product for carrying out one of the above methods. Still further, the present invention also provides a postfilter for attenuating inter-harmonic noise, operable in an active mode or a pass-through mode as indicated by a post-filtering signal supplied to the post-filter, the post-filter optionally having a decision unit for autonomously controlling the post-filtering operation.

[0018] Those skilled in the art will appreciate that an encoder configured to cooperate with a decoder is equipped with functionally equivalent modules, allowing faithful reproduction of the encoded signal. Such equivalent modules are identical or similar modules or modules having the same or similar transfer characteristics. In particular, the encoder and decoder modules may each be similar or dissimilar units executing computer programs that perform equivalent mathematical operations.

[0019] In one embodiment, the encoding of the method includes a decision on a postfilter that further comprises attenuation of the spectral valleys (of the formant envelope (see above)), which corresponds to the short-term part of the postfilter. It is advantageous to adapt the criteria on which the decision is based to the properties of the postfilter.

[0020] One embodiment relates to an encoder particularly adapted for speech coding. Since the problem motivating the present invention is found when encoding a mixture of vocals and other components, the combination of speech coding and independent decision-making regarding post-filtering according to the present invention is particularly advantageous. In particular, such a decoder may include a CELP coding module.

[0021] In one embodiment, the encoder's decision is based on detecting the co-occurrence of a signal component with a dominant fundamental frequency (pitch) with other signal components below the fundamental frequency. The detection may also aim to find the co-occurrence of a component with a dominant fundamental frequency with other components that have energy among the harmonics of this fundamental frequency. This is where artifacts of the type in question frequently occur. Thus, if such co-occurrence is found, the encoder determines that post-filtering is not appropriate. This is indicated by post-filtering information included in the bitstream.

[0022] In one embodiment, the detection criterion is the total signal power content in the audio time signal below the pitch frequency, possibly estimated by the encoder's long-term prediction. When this is greater than a predefined threshold, it is assumed that components other than pitch components (including harmonics) are present. Therefore, the post-filter is disabled.

[0023] In an encoder with a CELP module, such module can be used to estimate the pitch frequency of the audio time signal, and, as detailed above, a further detection criterion is to check the energy content between or below the harmonics of this frequency.

[0024] As a further development of the above embodiment including a CELP module, the decision may include a comparison between the estimated power of the audio signal when CELP coded (i.e., encoded and decoded) and the estimated power of the audio signal when CELP coded and post-filtered. If the power difference is greater than a threshold, this indicates that non-noise components of the signal are being lost, and the encoder decides to disable the post-filter.

[0025] In an advantageous embodiment, the encoder includes a CELP module and a TCX module. As is known in the art, TCX coding is advantageous for certain signals, particularly non-vocal signals. It is not common practice to apply post-filtering to TCX-coded signals. Thus, the encoder can select between TCX coding, CELP coding with post-filtering, or CELP coding without post-filtering to cover a wide range of signal types.

[0026] As a further development of the previous embodiment, the decision between the three coding modes is made based on a rate-distortion criterion, i.e. by applying an optimization procedure known per se in the art.

[0027] In another further development of the aforementioned embodiment, the encoder further comprises an AAC (Advanced Audio Coding) coder. AAC coders are also known to be particularly suitable for certain types of signals. Preferably, the decision to apply AAC (frequency domain) coding is made separately from the decision as to which other (linear prediction) mode to use. In this way, the encoder can be considered to operate in two super-modes: AAC or TCX CELP, where in the latter the encoder chooses between TCX, CELP with post-filtering, or CELP without filtering. This embodiment allows the processing of an even wider range of audio signal types.

[0028] In one embodiment, the encoder can decide to gradually apply post-filtering during decoding, i.e., gradually increase the gain. Similarly, the encoder can decide to gradually remove post-filtering. This gradual application and removal makes the switch between post-filtering and non-post-filtering less perceptible. As an example, a vocal section for which CELP coding with post-filtering is appropriate may be preceded by an instrumental section for which TCX coding is optimal. A decoder according to the present invention can gradually apply post-filtering at or near the beginning of the vocal section, thereby preserving the benefits of post-filtering while avoiding annoying switching artifacts.

[0029] In one embodiment, the decision to apply post-filtering is based on an approximated difference signal, which approximates the signal components that the post-filter will remove from the decoded signal. As an option, the approximated difference signal is calculated as the difference between the audio time signal and a (simulated) post-filtered audio time signal. As another option, the encoder retrieves an intermediate decoded signal, which allows the approximated difference signal to be calculated as the difference between the audio time signal and the post-filtered intermediate decoded signal. The intermediate decoded signal can be stored in the encoder's long-term prediction buffer, which represents the signal excitation and indicates the need for further synthesis filtering (vocal tracts, resonances) to obtain the final decoded signal. The point of using intermediate decoded signals is that they capture some of the peculiarities of the coding method, particularly its weaknesses, which allows for a more realistic estimation of the effect of the post-filter. As a third option, the decoder retrieves an intermediate decoded signal, which allows the approximated difference signal to be calculated as the difference between the intermediate decoded signal and the post-filtered intermediate decoded signal. The estimation by this procedure is probably less reliable than that by the first two options, but on the other hand it can be performed stand-alone by the decoder.

[0030] The approximated difference signal obtained in this way satisfies the following criteria: a) the power of the approximated difference signal exceeds a predetermined threshold and most of the signal is removed by the post-filter; b) whether the characteristics of the approximated difference signal are tonal rather than noise-like; c) whether the difference between the intensity frequency spectrum of the approximation difference signal and the intensity frequency spectrum of the audio time signal is unevenly distributed in frequency and is not noise but a signal that makes sense to a human listener; d) the intensity frequency spectrum of the approximated difference signal lies in a frequency interval within a predetermined relevant envelope that can be expected from a signal of the type being processed; e) whether the intensity frequency spectrum of the approximation difference signal lies in a frequency interval within a relevant envelope obtained by thresholding the intensity frequency spectrum of the audio time signal by the intensity of the maximum signal component downscaled by a predetermined scale factor; If it evaluates positively, a decision is made to disable the post filter.

[0031] When assessing criterion e), it is advantageous to apply peak tracking to the intensity spectrum, i.e. to distinguish those parts with peak-like shapes that are usually associated with tonal components rather than noise. The components identified by peak tracking, performed by any algorithm known per se in the art, are further sorted by applying a threshold to the peak heights. The remaining components are tonal material with a certain intensity. Such components usually represent relevant signal content rather than noise, which motivates the decision to disable the post-filter.

[0032] In one embodiment of the present invention as a decoder, the decision to disable the post-filter is performed by a switch controllable by the controller and allowing the post-filter to be bypassed in the circuit. In another embodiment, the post-filter has a variable gain controllable by the controller or by a gain controller therein, and the decision to disable is made by setting the post-filter gain (see previous section) to zero or by setting its absolute value below a pre-defined threshold.

[0033] In one embodiment, decoding according to the invention comprises retrieving post-filtering information from the bitstream signal to be decoded. More precisely, the post-filtering information is encoded in a data field comprising at least one bit in a format suitable for transmission. Advantageously, the data field is an existing field defined but unused by the applicable standard, and the post-filtering information does not increase the payload to be transmitted.

[0034] It should be noted that the methods and apparatus disclosed in this section can be applied to coding of signals having multiple components corresponding to different channels, such as stereo channels, with appropriate modifications within the ability of one skilled in the art, including routine experimentation. Throughout this application, pitch enhancement and post-filtering are used synonymously. It should also be noted that AAC is described as a representative example of a frequency-domain coding method. Indeed, application of the present invention to decoders and encoders operable in frequency-domain coding modes other than AAC requires only minor modifications, if any, within the ability of one skilled in the art. Similarly, TCX is described as an example of weighted linear predictive transform coding and transform coding in general.

[0035] In yet another embodiment, features of two or more of the above embodiments may be combined, unless they are clearly complementary. The fact that two features are recited in different claims does not mean that they cannot be combined. Similarly, in yet another embodiment, features that are not necessary or essential for the desired purpose may be omitted. [Brief explanation of the drawings]

[0036] Embodiments of the present invention will now be described in detail with reference to the accompanying drawings. [Figure 1] FIG. 1 is a block diagram illustrating a conventional decoder with a postfilter. [Figure 2] 1 is a block diagram illustrating a conventional decoder operable in AAC, ACELP, and TCX modes and including a post-filter connected downstream of the ACELP module; [Figure 3] FIG. 2 is a block diagram showing the configuration of a post filter. [Figure 4] FIG. 2 is a block diagram illustrating a decoder according to the present invention; [Figure 5] FIG. 2 is a block diagram illustrating another decoder according to the present invention. [Figure 6] FIG. 1 is a block diagram showing a conventional decoder. [Figure 7] 6 and 7 are block diagrams showing a decoder according to the present invention, and show the differences between a conventional decoder and a decoder according to the present invention. [Figure 8] 1 is a block diagram illustrating an encoder according to the present invention; [Figure 9] FIG. 1 is a block diagram showing a conventional decoder. [Figure 10] 9 and 10 are block diagrams showing a decoder according to the present invention, and show the differences between a conventional decoder and a decoder according to the present invention. [Figure 11] FIG. 2 is a block diagram illustrating an autonomous postfilter that can be selectively activated and deactivated. DETAILED DESCRIPTION OF THE INVENTION

[0037] FIG. 4 illustrates a decoder system 400 according to an embodiment of the present invention, which receives a bitstream signal and outputs an audio signal. Like the conventional decoder shown in FIG. 1, a postfilter 440 is placed after the decoding module 410 and can be added or removed from the decoding path by operating a switch 442. The postfilter is enabled in the switch position shown. The postfilter is disabled when the switch is set to the opposite position, thereby routing the signal from the decoding module 410 to a bypass line 444. As an inventive contribution, the switch 442 can be controlled by postfiltering information contained in the bitstream signal, allowing postfiltering to be applied or removed regardless of the current state of the decoding module 410. Because the postfilter 440 operates with a slight delay (e.g., the postfilter shown in FIG. 3 introduces a delay of at least the pitch period T), a compensating delay module 443 is placed in the bypass line 444 to keep the modules synchronized during switching. The delay module 443 delays the signal by the same amount as the postfilter 440, but does not process the signal. To minimize switching time, the compensation delay module 443 always receives the same signal as the postfilter 440. In embodiments where the postfilter 440 is replaced with a zero-delay postfilter (e.g., a causal filter with two taps that does not depend on future signal values), the compensation delay module 443 can be omitted.

[0038] FIG. 5 further develops the inventive teachings of the triple-mode decoder system 500 of FIG. 2. An ACELP decoding module 511 is arranged in parallel with a TCX decoding module 512 and an AAC decoding module 513. A post-filter 540 is arranged in series with the ACELP decoding module 511 to attenuate noise. Specifically, the post-filter 540 attenuates noise between harmonics of the pitch frequency, which is determined directly or indirectly from the bitstream signal to which the decoder system 500 is tuned. The bitstream signal also encodes post-filtering information that controls the position of an upper switch 541, as shown in FIG. 4. The upper switch 541 operates to remove the post-filter 540 from the processing path and replace it with a compensation delay 543. A lower switch 542 is used to switch between different decoding modes. In this configuration, the position of the upper switch 541 is not important when using either the TCX module 512 or the AAC module 513. Thus, the post-filtering information does not necessarily indicate this position, except in ACELP mode. Whatever decoding mode is currently in use, the signal is fed from the downstream connection of the lower switch 542 to the spectral band replication (SBR) module 550, which outputs the audio signal. Those skilled in the art will appreciate that the diagram is conceptual, and that the switches shown are shown as physically separate entities with movable contacts. In an actual embodiment of the decoder system, the switches and other modules would be implemented by computer-readable instructions.

[0039] 6 and 7 are also block diagrams illustrating a triple-mode decoder system operable in ACELP, TCX, or frequency-domain decoding modes. Referring to FIG. 7, one embodiment of the present invention is shown, in which a beamstream signal is fed to an input point 701. The input point 701 is permanently connected to decoding modules 711, 712, and 713 via respective branches. The input point 701 also has a connection branch 702 to a pitch enhancement module 740 (which is absent in the conventional decoding system of FIG. 6). This pitch enhancement module 740 functions as a postfilter of the general type described above. As is common practice in the art, a first transition window module 703 is located downstream of the ACELP module 711 and the TCX module 712 to transition between the decoding modules. A second transition module 704 is located downstream of the frequency-domain decoding module 713 and the first transition window module 703 to transition between the two super-modes. Furthermore, an SBR module 750 is provided immediately upstream of the output point 705. Obviously, the bitstream signal is directly (or, in some cases, after demultiplexing) supplied to all decoding modules 711, 712, 713 and the pitch enhancement module 740. Information contained in the bitstream controls which decoding modules are activated. However, in accordance with the present invention, the pitch enhancement module 740 is similarly self-activated. It functions as a postfilter or simply a pass-through module depending on the post-filtering information in the bitstream. This can be achieved, for example, by providing a control unit (not shown) in the pitch enhancement module 740 that turns on and off the post-filtering function. When the decoder system operates in frequency-domain decoding mode or TCX decoding mode, strictly speaking, post-filtering information is not required, and the pitch enhancement module 740 is always in pass-through mode. Needless to say, modules that do not constitute a contribution of the present invention and whose presence would be obvious to one skilled in the art, such as the demultiplexer, have been omitted from FIG. 7 and other similar figures for clarity.

[0040] 7 may also include a control module (not shown) that determines whether to apply post-filtering using an analysis-by-synthesis approach. Such a control module is communicatively coupled to the pitch enhancement module 740 and the ACELP module 711, and receives an intermediate decoded signal s representing an intermediate stage of the decoding process, preferably one that corresponds to the signal excitation. i_DEC (n). The detection module extracts the transfer function P LT (z) and H LP (z) (see Background and Figure 3), or equivalently, these filter impulse responses p LT (z) and h LP (n) has the information necessary to simulate the function of the pitch enhancement module 740 defined in (n). As explained in the Background section, the components subtracted in post-filtering are [Outside 1] Approximate difference signal s proportional to TIFF2025183321000005.tif7115 AD (n), where * denotes discrete convolution. This is an approximation of the true difference between the original audio signal and the post-filtered decoded signal. That is,

number

[0041] FIG. 8 illustrates an encoder system 800 according to an embodiment of the present invention. The encoder system 800 is configured to process digital audio signals, typically obtained by capturing sound waves with a microphone and converting them into analog electrical signals. The electrical signals are sampled to produce digital signals that can be provided to the encoder system 800 in a suitable format. The system generally comprises an encoding module 810, a decision module 820, and a multiplexer 830. Symbolically represented switches 814 and 815 allow the encoding module 810 to operate in either CELP, TCX, or AAC mode by selectively activating modules 811, 812, and 813. The decision module 820 applies predetermined criteria to determine whether to encode the audio signal with post-filtering disabled during decoding of the bitstream signal generated by the encoding system 800. For this purpose, the decision module 820 examines the audio signal directly or receives data from the encoding module 810 via a connection line 816. A signal indicative of the decision made by decision module 820, along with the encoded audio signal from encoding module 810, is provided to multiplexer 830, which combines these signals into a bitstream that constitutes the output of encoder system 800.

[0042] Preferably, the decision module 820 determines whether the intermediate decoded signal s i_DEC The decision is based on the approximate difference signal calculated from this intermediate decoded signal s i_DEC can be obtained from the encoding module 810. As mentioned above, the intermediate decoded signal represents an intermediate stage in the decoding process, but can also be obtained from a corresponding stage in the encoding process. However, in the encoder system 800, the approximate difference signal is advantageously

number

[0043] As a variation of this, the decision module 820 may use the original signal instead of the intermediate decoded signal to determine whether the approximate difference signal is [Outside 2] TIFF2025183321000008.tif8115. This is a less faithful approximation, but on the other hand the presence or absence of the connection line 816 between the decision module 820 and the encoding module 810 is optional.

[0044] In other variations of this embodiment where the decision module 820 studies the audio signal directly, one or more of the following criteria may be applied: Does the audio signal contain both fundamental frequency dominated components and components below the fundamental frequency? (The fundamental frequency is provided as a side effect of the encoding module 810.) · Does the audio signal contain both components dominated by the fundamental frequency and components that lie between the harmonics of the fundamental frequency? ·Does the audio signal contain significant signal energy below the fundamental frequency? Is post-filtered decoding (likely) preferable to unfiltered decoding in terms of rate-distortion optimality? In all the variations described for the encoder configuration shown in Figure 8, i.e. regardless of the underlying decision criterion, the decision unit 820 can decide to gradually apply or remove post-filtering in a way that achieves a smooth transition. Post-filtering can be gradually applied or removed by adjusting the post-filter gain.

[0045] FIG. 9 shows a conventional decoder that can operate in frequency-domain and CELP decoding modes depending on the bitstream signal provided to the decoder. Post-filtering is applied whenever CELP decoding mode is selected. An improvement to this decoder is shown in FIG. 10. FIG. 10 shows a decoder 1000 according to one embodiment of the present invention. This decoder is operable not only in a frequency-domain based decoding mode in which the frequency-domain decoding module 1013 is active, and a filtered CELP decoding mode in which the CELP decoding module 1011 and post-filter 1040 are active, but also in an unfiltered CELP mode in which the CELP module 1011 provides its signal to the compensation delay module 1043 via a bypass line 1044. A switch 1042 controls which decoding mode is currently used depending on the post-filtering information included in the bitstream signal provided to the decoder 1000. In this decoder and the decoder of FIG. 9, the final processing step is performed by the SBR module 1050, which outputs the final audio signal.

[0046] FIG. 11 illustrates a postfilter 1100 suitable for placement downstream of a decoder 1199. The filter 1100 includes a postfiltering module 1140, which is enabled or disabled by a control module (not shown), specifically a binary or non-binary gain controller, in response to a postfiltering signal received from a decision module 1120 within the postfilter 1100. The decision module performs one or more tests on the signal obtained from the decoder to determine whether to activate the postfiltering module 1140. This decision is made at a point in time similar to the function of decision module 820 in FIG. 8. The decision module 820 uses the original signal and / or intermediate decoded signals to predict the operation of the postfilter. Furthermore, the decision module 1120's decisions are based on information similar to that used in the above-described embodiments in which the intermediate decoded signals are constructed. As an example, the decision module 1120 estimates the pitch frequency (unless the pitch frequency is readily available from the bitstream signal) and calculates the energy content of the signal below the pitch frequency and between its harmonics. If this energy content is large, it will represent a signal component rather than noise, which motivates the decision to disable the post-filtering module 1140.

[0047] A listening test was conducted with six subjects, during which music samples coded and decoded according to the present invention were compared with a reference sample containing the same music coded. Post-filtering was performed as usual, but all other parameters were left unchanged. The resulting improvement in sound quality was confirmed.

[0048] Further embodiments of the present invention will be apparent to those skilled in the art upon reading the above description. While the specification and drawings disclose embodiments and examples, the present invention is not limited to these specific examples. Numerous modifications and variations can be made without departing from the scope of the present invention, which is defined in the appended claims.

[0049] The systems and methods disclosed herein can be implemented as software, firmware, hardware, or a combination thereof. Some or all of the components can be implemented as software executed by a digital signal processor or microprocessor, or as hardware or special-purpose integrated circuits. Such software can be distributed on computer-readable media. Computer-readable media includes computer storage media (i.e., non-transitory media) and communication media (i.e., transitory media). As known to those skilled in the art, computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory and other memory technology, CD-ROM, Digital Versatile Disk (DVD) and other optical disk storage media, magnetic cassettes, magnetic tape, magnetic disk storage and other magnetic storage devices, or any other medium capable of storing the desired information. Additionally, as known to those skilled in the art, communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.

[0050] List of Embodiments 1. A decoder system (400; 500; 700; 1000) for decoding a bitstream signal as an audio time signal, comprising: a decoding unit (410; 511, 512, 513; 711, 712, 713; 1011, 1013) for decoding the bitstream signal as a preliminary audio time signal; an inter-harmonic noise attenuating post-filter (440; 540; 740; 1040) for filtering the preliminary audio time signal to obtain an audio time signal; a control unit configured to disable the post-filter in response to post-filtering information encoded in the bitstream signal; 2. A decoder system, wherein said preliminary audio time signal is output as said audio time signal.

[0051] 2. The decoder system of embodiment 1, wherein the post-filter is further configured to attenuate noise in spectral valleys.

[0052] 3. The decoder system of embodiment 1, wherein the control unit includes a switch (442; 541; 1042) that selectively removes the post-filter from the signal processing path of the decoder system, thereby disabling the post-filter.

[0053] 4. The decoder system of embodiment 1, wherein the post-filter has a variable gain that determines the inter-harmonic attenuation, and the control unit includes a gain controller operable to set the absolute value of the gain below a predetermined threshold.

[0054] 5. The decoder system of embodiment 1, wherein the decoding unit includes a speech decoding module.

[0055] 6. The decoder system of embodiment 1, wherein the decoding unit includes a Code Excited Linear Prediction, CELP, decoding module (511; 711; 1011).

[0056] 7. A decoder system as described in embodiment 5, wherein a pitch frequency estimated by a long-term prediction unit of the encoder is encoded into the bitstream signal.

[0057] 8. The decoder system of embodiment 7, wherein the postfilter is configured to attenuate spectral components that lie between harmonics of the pitch frequency.

[0058] 9. The bitstream signal includes a representation of a pitch frequency, and the postfilter is configured to attenuate spectral components that lie between harmonics of the pitch frequency. 2. A decoder system as described in embodiment 1.

[0059] 10. A decoder system as described in embodiment 8 or 9, wherein the post-filter is configured to attenuate only spectral components below a predetermined cut-off frequency.

[0060] 11. The decoding unit further comprises a TCX (transform-coded excitation) decoding module (512; 712) for decoding the bitstream signal as an audio time signal; The control unit has at least the following modes: a) the TCX module is enabled and the post filter is disabled; b) the CELP module and the post-filter are enabled; and c) the CELP module is enabled, the post filter is disabled, and the call audio time signal and the audio time signal are simultaneous; 7. The decoder system of embodiment 6, configured to operate the decoder system at

[0061] 12. The decoding unit further includes an AAC (Advanced Audio Coding) decoding module (513; 713) for decoding the bitstream signal as an audio time signal; The control unit has at least the following modes: d) The AAC module is enabled and the post filter is disabled. 11. The decoder system of embodiment 10, configured to operate the decoder system at

[0062] 13. The bitstream signal is divided into time frames, and the control unit is configured to disable an entire time frame or a series of time frames. 2. A decoder system as described in embodiment 1. 14. The decoder system of embodiment 13, wherein the control unit is further configured to receive, for each time frame in an MPEG (Moving Pictures Experts Group) bitstream, a data field associated with the time frame, and is operable to disable the post filter depending on the value of the data field.

[0063] 15. A decoder system as described in embodiment 4, wherein the control unit is configured to gradually decrease and / or increase the gain of the post filter.

[0064] 16. A decoder system (400; 500; 700; 1000) comprising: a decoding unit (410; 511, 512, 513; 711, 712, 713; 1011, 1013) for decoding the bitstream signal as a preliminary audio time signal; an inter-harmonic noise attenuating post-filter (440; 540; 740; 1040) for filtering the preliminary audio time signal to obtain an audio time signal; the decoding unit is configured to generate an intermediate decoded signal representative of the excitation and provide it to the control unit; The control unit performs the post-filtering on the intermediate decoded signal and filters the intermediate decoded signal based on the following criteria: a) the power of the approximated difference signal exceeds a predetermined threshold; b) the characteristics of said approximated difference signal are tonal; c) the difference between the magnitude-frequency spectrum of the approximation difference signal and the magnitude-frequency spectrum of the audio signal is non-uniformly distributed with respect to frequency; d) the intensity frequency spectrum of the approximated difference signal is in a frequency interval within a predetermined associated envelope; e) whether the intensity frequency spectrum of the approximation difference signal lies in a frequency interval within a relevant envelope obtained by thresholding the intensity frequency spectrum of the audio time signal by the intensity of the maximum signal component downscaled by a predetermined scale factor; and calculating an approximate difference signal that approximates the signal component removed from the decoded signal by the post-filter as a difference between the intermediate decoded signal and the intermediate decoded signal evaluated by at least one of the following: In response to a positive determination, the decoder system disables the post-filter, whereby the preliminary audio time signal is output as the audio time signal.

[0065] 17. An inter-harmonic noise attenuating post-filter (440; 550; 740; 1040; 1140) configured to receive an input signal including a preliminary audio signal and to provide an output audio signal, comprising: Depending on the value of the post-filtering signal, the following modes are available: i) a filtering mode in which the preliminary audio signal is filtered to obtain a filtered signal which is provided as an output audio signal; ii) a pass-through mode in which the preliminary audio signal is provided as the output audio signal; 1. A post-filter comprising: a control unit for selectively operating the post-filter;

[0066] 18. The post-filtering signal is included in the input signal. 18. A post filter according to embodiment 17.

[0067] 19. Estimate the pitch frequency of the preliminary audio signal according to the following criteria: a) whether the power of the spectral components below the pitch frequency exceeds a predetermined threshold; b) whether the spectral components below said pitch frequency are tonal; c) whether the power of the spectral components between harmonics of the pitch frequency exceeds a predetermined threshold; d) whether the spectral content between the harmonics of said pitch frequency is tonal; and a decision module (1120) configured to evaluate at least one of: In response to a positive determination, determining to generate a negative post-filtering signal that disables the post-filter. 18. A post filter according to embodiment 17.

[0068] 20. A method for decoding a bitstream signal as an audio time signal, comprising: decoding the bitstream signal as a preliminary audio time signal; post-filtering the preliminary audio time signal by attenuating inter-harmonic noise, thereby obtaining an audio time signal; 10. A method according to claim 9, wherein the step of post-filtering is selectively omitted depending on post-filtering information encoded in the bitstream signal.

[0069] 21. The method of embodiment 20, wherein the post-filtering step further comprises attenuating noise in spectral valleys.

[0070] 22. The method of embodiment 20, wherein the decoding step includes applying a coding method for speech coding.

[0071] 23. The method of embodiment 20, wherein the decoding step includes applying code-excited linear prediction (CELP) decoding.

[0072] 24. The post-filtering step includes attenuating spectral components between harmonics of the pitch frequency, the pitch frequency being derived from the bitstream signal or estimated in the decoding step. 24. The method of embodiment 22 or 23.

[0073] 25. The post-filtering step includes attenuating only spectral components below a predetermined cutoff frequency. 21. The method of embodiment 20.

[0074] 26. The decoding and post-filtering steps include: a)TCX decoding; b) CELP decoding with post-filtering; and c) CELP decoding without post-filtering; 24. The method of embodiment 23, wherein selectively performing one of:

[0075] 27. The decoding and post-filtering steps are performed in modes a), b), c), and d) AAC (Advanced Audio Coding) decoding 27. The method of embodiment 26, wherein selectively performing one of:

[0076] 28. The bitstream signal is divided into time frames, and the post-filtering step is omitted for an entire time frame or a series of time frames. 21. The method of embodiment 20.

[0077] 29. The bitstream signal is an MPEG (Moving Pictures Experts Group) bitstream and includes, for each time frame, an associated data field; 29. The method of embodiment 28, wherein the post-filtering step is omitted in a time frame depending on the value of the associated data field.

[0078] 30. The post-filtering omission is as follows: Complete omission of attenuation, Partial omission of attenuation, Gradually increasing the attenuation, and Gradually decreasing the attenuation, 21. The method of embodiment 20, including one of

[0079] 31. A method for decoding a bitstream signal as an audio time signal, comprising: decoding the bitstream signal as a preliminary audio time signal; post-filtering the preliminary audio time signal by attenuating inter-harmonic noise, thereby obtaining an audio time signal; The step of decrypting includes: deriving an intermediate decoded signal representative of the excitation; calculating an approximate difference signal that approximates the signal components removed from the decoded signal by the post-filter as a difference between the intermediate decoded signal and a post-filtered intermediate decoded signal; The following criteria: a) the power of the approximated difference signal exceeds a predetermined threshold; b) the characteristics of said approximated difference signal are tonal; c) the difference between the magnitude-frequency spectrum of the approximation difference signal and the magnitude-frequency spectrum of the audio signal is non-uniformly distributed with respect to frequency; d) the intensity frequency spectrum of the approximated difference signal is in a frequency interval within a predetermined associated envelope; e) whether the intensity frequency spectrum of the approximation difference signal lies in a frequency interval within a relevant envelope obtained by thresholding the intensity frequency spectrum of the audio time signal by the intensity of the maximum signal component downscaled by a predetermined scale factor; evaluating at least one of: In response to a positive determination, the encoder system disables the post-filter, whereby the preliminary audio time signal is output as the audio time signal.

[0080] 32. An encoder system (800) for encoding an audio time signal as a bitstream signal, comprising: An encoding unit (810) for encoding an audio time signal as a bitstream signal, a decision unit (820) configured to determine whether to disable post-filtering including attenuation of inter-harmonic noise in decoding the bitstream, and to encode this decision as post-filtering information into the bitstream signal.

[0081] 33. The decision unit is configured to decide whether to disable post-filtering, further including attenuation of noise in spectral valleys. An encoder system as described in embodiment 32.

[0082] 34. The encoding unit includes a speech coding module. An encoder system as described in embodiment 32.

[0083] 35. An encoder system as described in embodiment 32, wherein the encoding unit includes a code-excited linear prediction (CELP) encoding module.

[0084] 36. The decision unit: detecting the presence of both a signal component having a dominant fundamental frequency and signal components below said fundamental frequency and, optionally, among its harmonics; and make the decision to disable accordingly. An encoder system as described in embodiment 32.

[0085] 37. The CELP encoding module is configured to estimate a pitch frequency in the audio time signal; the decision unit is configured to detect spectral components below the estimated pitch frequency and make a disabling decision accordingly. An encoder system as described in embodiment 35.

[0086] 38. The decision unit: calculating a difference between a predicted power of the audio time signal when CELP coded and a predicted power of the audio time signal when CELP coded and post-filtered; making a decision to disable in response to the difference exceeding a predetermined threshold; An encoder system as described in embodiment 35.

[0087] 39. The encoding unit further includes a transform-coded excitation (TCX) encoding module; The determining unit is configured to determine the following encoding modes: a)TCX encoding; b) CELP coding with post-filtering; and c) CELP coding without post-filtering; 36. The encoder system of embodiment 35, configured to select one of:

[0088] 40. Next Super Mode: i) AAC (Advanced Audio Coding) coding in which the decision unit is disabled; ii) TCX / CELP coding, wherein the decision unit can select one of coding modes a), b) and c); 39. The encoder system of claim 39, further comprising a coding selector (814) configured to select one of:

[0089] 41. An encoder system as described in embodiment 39, wherein the decision unit is configured to decide which mode to use based on rate-distortion optimization.

[0090] 42. An encoder system as described in embodiment 32, configured to divide the bitstream signal into time frames, and wherein the decision unit is configured to make a decision to disable the post filter in a time segment consisting of an entire frame.

[0091] 43. An encoder system as described in embodiment 32, wherein the decision unit is configured to decide to gradually decrease and / or increase the attenuation of the post filter.

[0092] 44. The decision unit: Calculate the power of the audio time signal below the estimated pitch frequency; making a disable decision in response to this power exceeding a predetermined threshold; An encoder system as described in embodiment 32.

[0093] 45. The decision unit: determining from the audio time signal an approximation difference signal that approximates the signal components that are removed from the further decoded signal by the post-filter; The following criteria: a) the power of the approximated difference signal exceeds a predetermined threshold; b) the characteristics of said approximated difference signal are tonal; c) the difference between the magnitude-frequency spectrum of the approximation difference signal and the magnitude-frequency spectrum of the audio signal is non-uniformly distributed with respect to frequency; d) the intensity frequency spectrum of the approximated difference signal is in a frequency interval within a predetermined associated envelope; e) whether the intensity frequency spectrum of the approximation difference signal lies in a frequency interval within a relevant envelope obtained by thresholding the intensity frequency spectrum of the audio time signal by the intensity of the maximum signal component downscaled by a predetermined scale factor; Evaluate at least one of the following: in response to a positive determination, determining to disable the post-filter; An encoder system as described in embodiment 32.

[0094] 46. ​​The decision unit is configured to calculate the approximation difference signal as the difference between the audio time signal and a post-filtered audio time signal. An encoder system as described in embodiment 45.

[0095] 47. The encoding unit is configured to retrieve an intermediate decoded signal representing the excitation and provide it to the decision unit; the determination unit is configured to calculate the approximation difference signal as the difference between the audio time signal and a post-filtered intermediate decoded signal. An encoder system as described in embodiment 45.

[0096] 48. A method of encoding an audio time signal as a bitstream signal, comprising the steps of: encoding the audio time signal as a bitstream signal; 10. The method of claim 9, further comprising the step of: determining whether to disable post-filtering, including attenuation of inter-harmonic noise, in decoding the bitstream; and encoding this decision as post-filtering information into the bitstream signal.

[0097] 49. The method of embodiment 48, wherein the determining step further relates to post-filtering, which includes attenuating noise in spectral valleys.

[0098] 50. The method of embodiment 48, wherein the encoding step includes applying a coding method for speech coding.

[0099] 51. The method of embodiment 48, wherein the encoding step includes applying code-excited linear prediction (CELP) coding.

[0100] 52. The method according to claim 1, further comprising the step of detecting the presence of both a signal component having a dominant fundamental frequency and signal components below said fundamental frequency and, optionally, among its harmonics; In the case of a positive decision, a decision is made to disable post-filtering. 49. The method of embodiment 48.

[0101] 53. The step of CELP coding includes a step of estimating a pitch frequency in the audio time signal; the determining step comprises detecting spectral components below the estimated pitch frequency, and in case of a positive detection result, a decision is made to disable post-filtering. 52. The method of embodiment 51.

[0102] 54. Further comprising the step of calculating a difference between the predicted power of the audio time signal when CELP coded and the predicted power of the audio time signal when CELP coded and post-filtered; 52. The method of embodiment 51, wherein if this difference exceeds a predetermined threshold, a decision is made to disable post-filtering.

[0103] 55. The encoding step includes selectively applying either CELP coding or TCX coding; The step of determining whether to disable post-filtering is performed only when CELP coding is applied. 52. The method of embodiment 51. 56. The determining step is based on rate-distortion optimization and includes determining the following operating modes: a)TCX coding; b) CELP coding with post-filtering; and c) CELP coding without post-filtering; 56. The method of embodiment 55, comprising selecting one of:

[0104] 57. The determining step is based on rate-distortion optimization and includes determining the following operating modes: a)TCX coding; b) CELP coding with post-filtering; and c) CELP coding without post-filtering; and d) AAC (Advanced Audio Coding) coding 56. The method of embodiment 55, comprising selecting one of:

[0105] 58. The encoding step includes dividing the audio time signal into time frames and constructing a bitstream signal having corresponding time frames; the step of determining to disable post-filtering occurs once per time frame; 49. The method of embodiment 48.

[0106] 59. The step that decides to disable post-filtering results from: No attenuation, Total attenuation, partial attenuation, Gradually increasing the attenuation, and Gradually decreasing the attenuation, 49. The method of embodiment 48, wherein said method is selected.

[0107] 60. The determining step includes calculating the power of the audio time signal below the estimated pitch frequency, and disabling the post-filter in response to the power exceeding a predetermined threshold. 49. The method of embodiment 48.

[0108] 61. The encoding step includes a step of obtaining an approximation difference signal from the audio time signal that approximates a signal component that is removed from the further decoded signal by the post-filter. The decision step is based on the following criteria: a) the power of the approximated difference signal exceeds a predetermined threshold; b) the characteristics of said approximated difference signal are tonal; c) the difference between the magnitude-frequency spectrum of the approximation difference signal and the magnitude-frequency spectrum of the audio signal is non-uniformly distributed with respect to frequency; d) the intensity frequency spectrum of the approximated difference signal is in a frequency interval within a predetermined associated envelope; e) whether the intensity frequency spectrum of the approximation difference signal lies in a frequency interval within a relevant envelope obtained by thresholding the intensity frequency spectrum of the audio time signal by the intensity of the maximum signal component downscaled by a predetermined scale factor; and and disabling said post-filter in response to at least a positive determination. 49. The method of embodiment 48.

[0109] 62. Calculating the approximate difference signal as the difference between the audio time signal and a post-filtered audio time signal. 62. The method of embodiment 61.

[0110] 63. The encoding step includes the step of deriving an intermediate decoded signal representing the excitation; the determining step comprises calculating the approximation difference signal as the difference between the audio time signal and a post-filtered intermediate decoded signal. 62. The method of embodiment 61.

[0111] 64. A computer-readable medium including a data carrier having stored thereon instructions for carrying out the method according to any one of embodiments 20 to 31 and 48 to 63.

Claims

1. 1. A method for decoding a bitstream signal as an audio time signal, comprising: decoding the bitstream signal as a preliminary audio time signal according to a coding mode selected from a plurality of coding modes, the plurality of coding modes including at least a first coding mode including a post-filtering step and at least a second coding mode not including a post-filtering step; the post-filtering step includes applying a pitch enhancement filter to the preliminary audio time signal to obtain an audio time signal, wherein the pitch enhancement is reduced for higher frequencies; the post-filtering step is selectively omitted in response to post-filtering information encoded in the bitstream signal, the post-filtering information indicating an encoder-side decision of whether to omit the post-filtering step, whereby the post-filtering step is selectively omitted in the first coding mode. method.

2. The method of claim 1 , wherein the higher frequency is above a predetermined cutoff frequency.

3. 2. The method of claim 1, wherein the decoding step includes applying code-excited linear prediction (CELP) decoding.

4. The method of claim 1 , wherein the bitstream signal is divided into time frames, and the post-filtering step is omitted for an entire time frame or a sequence of entire time frames.

5. A computer program product comprising a data carrier having stored thereon instructions which, when executed by a digital signal processor, cause said digital signal processor to perform a method according to any one of claims 1 to 4.

6. A decoding system configured to perform the method according to any one of claims 1 to 4.

Images