RU2769429C2 - Audio signal encoder - Google Patents

Abstract

FIELD: computer equipment for encoding audio data.

SUBSTANCE: technical result is achieved by determining the distance between the potential code vector and the ordered vector of parameters by forming the first and second distance components based on the first value of the scaling factor, wherein the first and second distance components are associated with the distance between the potential code vector and the ordered parameter vector, correcting the sign of the first distance component, when the potential code vector does not pass the parity constraint check, and determining the distance between the at least one potential code vector and the ordered parameter vector based on the first and second distance components; determining the leading class, associated with potential code vector, which forms the least associated distance; and sorting the leading class by descending ordering based on the absolute values of the components of said at least one parameter vector.

EFFECT: reduced complexity of encoding by estimating distortion of quantization in transposed vector space.

15 cl, 6 dwg

Description

FIELD OF TECHNOLOGY

The present application relates to a multi-channel or stereo audio encoder, and in particular, but not exclusively, to a multi-channel or stereo audio encoder for use in a portable device.

BACKGROUND OF THE INVENTION

Audio signals such as speech or music are encoded, for example, to enable efficient transmission or storage.

Audio encoders and decoders (also referred to as codecs) are used to represent audio-based signals such as music and audio (which may be referred to as background noise in speech coding terms). In these types of codecs, the speech reproduction model is usually not used for the coding process; to a greater extent, these codecs use the processes for representing audio signals of all types, including the speech signal. Speech encoders and decoders (codecs) can be thought of as audio codecs that are optimized for speech signals and can operate at either a fixed or variable bit rate.

Audio encoders and decoders are often designed as simple source codecs. In other words, these devices are capable of encoding and decoding audio signals without the need for complicated processing.

An example is coding with a transformation. In the case of a music signal, transform coding of an audio signal is performed more efficiently than Algebraic Code Excited Linear Prediction (ACELP) technology, which is better adapted and directed to speech signal processing. Transform encoding is performed by transform encoding coefficients of subband vectors, in other words, the audio signal is divided into subbands for which a parameter is determined, and the parameters represent subvectors resulting from vector or lattice quantization.

SUMMARY OF THE INVENTION

In accordance with the first aspect of the present invention, a method is provided, including the formation of at least one parameter vector that defines at least one audio signal; sorting at least one parameter vector according to the order of at least one of the vector's absolute tuples to form at least one associated ordered parameter vector; selecting at least one potential code vector from the list of leading classes; determining a distance between at least one candidate code vector and at least one ordered parameter vector; determining at least one leading class associated with the potential code vector that generates the smallest associated distance; and transposing at least one leading class to generate an output code vector derived from the trellis quantization.

The method may also include selecting at least one scale factor; while determining the distance between at least one potential code vector and at least one ordered parameter vector may also include: determining the distance based on at least one scale factor; determining at least one leading class associated with the candidate code vector that generates the smallest associated distance may also include determining an output scale factor associated with the candidate code vector and the scale factor that generates the smallest associated distance; and transposing at least one leading class to generate an output trellis-quantized codevector may also include applying an output scaling factor to the output trellis-quantized codevector.

The formation of the first vector of parameters that define at least one audio signal may include: dividing at least one audio signal into time frames; and determining a vector of discrete spectrum frequency parameters associated with at least one of the time frames of the audio signal.

Sorting at least one parameter vector according to the order of at least one of the vector's absolute tuples to form at least one associated ordered parameter vector may include: determining weights to determine a weighted distance; sorting the weights based on the order of said at least one of the absolute tuples of the vector to form a vector of sorted weights; and applying the sorted weight vector to the at least one ordered parameter vector.

Determining the distance between at least one potential code vector and at least one ordered parameter vector may include: generating the first and second distance components based on the first scale factor value, wherein the first and second distance components are related to the distance between the at least one potential a code vector and at least one ordered parameter vector; correcting the sign of the first distance component when the candidate codevector fails the parity constraint test; and determining a distance between the at least one candidate code vector and the at least one ordered parameter vector based on the first and second distance components.

Determining a distance between at least one candidate code vector and at least one ordered parameter vector may also include: determining a distance between at least one candidate code vector and at least one ordered parameter vector based on the first and second distance components, and based on scale factor values.

Determining at least one leading class associated with a potential codevector that generates the smallest associated distance may include: determining an index of the potential codevector that generates the smallest associated distance; and determining a scale factor of the candidate codevector that generates the smallest associated distance by applying the scale factors only to the candidate codevector that generates the smallest associated distance.

Transposing the at least one leading class to generate an output codevector derived from the trellis quantization may include sorting the at least one leading class by reordering at least one of the vector's absolute tuples.

According to a second aspect, a device is provided, comprising means for generating at least one vector of parameters defining at least one audio signal; means for sorting the at least one parameter vector according to the order of at least one of the vector's absolute tuples to form at least one associated ordered parameter vector; means for selecting at least one potential codevector from a list of leading classes; means for determining a distance between at least one candidate code vector and at least one ordered parameter vector; means for determining at least one leading class associated with the potential code vector that generates the smallest associated distance; and means for transposing the at least one leading class to generate an output code vector derived from the trellis quantization.

The device may also include means for selecting at least one scale factor; the means for determining the distance between the at least one potential code vector and the at least one ordered parameter vector may also include: means for determining the distance based on the at least one scale factor; means for determining at least one leading class associated with the candidate code vector that generates the smallest associated distance may also comprise means for determining an output scale factor associated with the candidate code vector and the scale factor that generates the smallest associated distance; and means for transposing at least one leading class to generate an output codevector derived from the trellis quantization may also include means for applying an output scaling factor to the output codevector derived from the trellis quantization.

Means for generating a first vector of parameters defining at least one audio signal may include: means for separating at least one audio signal into time frames; and means for determining a vector of discrete spectrum frequency parameters associated with at least one of the time frames of the audio signal. Means for sorting at least one parameter vector according to the order of at least one of the absolute tuples of the vector to form at least one associated ordered parameter vector may comprise: means for determining weights for determining a weighted distance; means for sorting weights based on the order of said at least one of the absolute tuples of the vector to generate a vector of sorted weights; and means for applying the sorted weight vector to the at least one ordered parameter vector.

Means for determining the distance between at least one potential code vector and at least one ordered parameter vector may comprise: means for generating the first and second distance components based on the first scale factor value, wherein the first and second distance components are associated with the distance between at least one potential code vector and at least one ordered parameter vector; means for correcting the sign of the first distance component when the candidate codevector fails a parity constraint test; and means for determining a distance between the at least one candidate code vector and the at least one ordered parameter vector based on the first and second distance components.

The means for determining the distance between at least one potential code vector and at least one ordered parameter vector may also comprise means for determining the distance between at least one potential code vector and at least one ordered parameter vector based on the first and second distance components, and also based on the values of scale factors.

Means for determining at least one leading class associated with a potential codevector that generates the smallest associated distance may comprise: means for determining an index of the potential codevector that generates the smallest associated distance; and means for determining a scale factor of the candidate code vector that generates the smallest associated distance by applying the scale factor values only to the candidate code vector that generates the smallest associated distance.

Means for transposing at least one leading class to form an output code vector derived from the trellis quantization may comprise means for sorting at least one leading class by reordering at least one of the vector's absolute tuples.

In accordance with a third aspect, an apparatus is provided, comprising at least one processor and at least one memory containing computer program code, wherein at least one memory and computer program code are configured such that when interacting with at least one processor the device performed at least the following: generating at least one parameter vector defining at least one audio signal; sorting at least one parameter vector according to the order of at least one of the vector's absolute tuples to form at least one associated ordered parameter vector; selecting at least one potential code vector from the list of leading classes; determining a distance between at least one candidate code vector and at least one ordered parameter vector; determining at least one leading class associated with the potential code vector that generates the smallest associated distance; and transposing at least one leading class to generate an output code vector derived from the trellis quantization.

The device may also perform the following: selecting at least one scale factor; while determining the distance between at least one potential code vector and at least one ordered parameter vector may also include: determining the distance based on at least one scale factor; wherein determining at least one leading class associated with the candidate code vector that generates the smallest associated distance may also include determining an output scale factor associated with the candidate code vector and the scale factor that generates the smallest associated distance; wherein transposing at least one leading class to generate an output codevector derived from the trellis quantization may include applying an output scaling factor to the output codevector derived from the trellis quantization.

Forming the first vector of parameters defining at least one audio signal may include dividing at least one audio signal into time frames and determining a vector of discrete spectrum frequency parameters associated with at least one of the time frames of the audio signal.

Sorting at least one parameter vector according to the order of at least one of the absolute vector tuples to form at least one associated ordered parameter vector may include determining weights to determine a weighted distance, to sort the weights based on the order of said at least one of the absolute tuples a vector to generate a sorted weights vector; and applying the sorted weights vector to at least one ordered parameter vector.

Determining the distance between the at least one candidate code vector and the at least one ordered parameter vector may include generating first and second distance components based on the first scale factor value, wherein the first and second distance components are related to the distance between the at least one candidate code vector. vector and at least one ordered parameter vector, to correct the sign of the first distance component when the candidate code vector fails the parity constraint test, and determining the distance between the at least one candidate code vector and the at least one ordered parameter vector based on the first and the second distance component.

Determining the distance between the at least one candidate code vector and the at least one ordered parameter vector may also include determining the distance between the at least one candidate code vector and the at least one ordered parameter vector based on the first and second distance components, as well as on based on scale factor values.

Determining at least one leading class associated with the potential codevector that generates the smallest associated distance may include determining an index of the potential codevector that generates the smallest associated distance and determining a scale factor of the potential codevector that generates the smallest associated distance by applying scalefactors only to the potential codevector that generates the smallest associated distance.

Transposing at least one leading class to generate an output codevector derived from the trellis quantization may include performing a sort on at least one leading class by reordering at least one of the vector's absolute tuples.

According to a fourth aspect of the present invention, there is provided an apparatus comprising: a vector generator configured to generate at least one vector of parameters defining at least one audio signal; and a lattice vector quantizer configured to sort at least one parameter vector according to the order of at least one of the absolute tuples of the vector to form at least one associated ordered parameter vector, select at least one potential code vector from the list of leading classes, determine the distance between at least one candidate codevector and at least one ordered parameter vector, determining at least one leading class associated with the candidate codevector that generates the smallest associated distance, and transposing at least one leading class to form an output codevector , obtained on the basis of lattice quantization.

An apparatus wherein the trellis vector quantizer is also configured to select at least one scale factor; while the lattice vector quantizer, configured to determine the distance between at least one potential code vector and at least one ordered parameter vector, can also be configured to determine the distance based on at least one scale factor; a trellis vector quantizer configured to determine at least one leading class associated with the candidate code vector that generates the smallest associated distance may also be configured to determine an output scale factor associated with the candidate code vector and the scale factor that generates the smallest associated distance ; and a trellis vector quantizer configured to transpose at least one leading class to generate an output codevector derived from the trellis quantization may also be configured to apply an output scaling factor to the output codevector derived from the trellis quantization.

The apparatus may also comprise a parameter determiner configured to divide at least one audio signal into time frames and to determine a vector of discrete spectrum frequency parameters associated with at least one of the time frames of the audio signal.

At the same time, a lattice vector quantizer configured to sort at least one parameter vector according to the order of at least one of the absolute tuples of the vector to form at least one associated ordered parameter vector can also be configured to determine weights for determining a weighted distance, for sorting weights based on the order of said at least one of the absolute tuples of the vector to form a vector of sorted weights, and to apply the sorted weights vector to at least one ordered parameter vector.

A lattice vector quantizer configured to determine a distance between at least one potential code vector and at least one ordered parameter vector may also be configured to generate first and second distance components based on the first scale factor value, wherein the first and second distance components associated with the distance between at least one candidate codevector and at least one ordered parameter vector, to correct the sign of the first distance component when the candidate codevector fails a parity constraint test, and to determine the distance between the at least one candidate codevector and at least one ordered parameter vector based on the first and second distance components.

A lattice vector quantizer configured to determine a distance between at least one candidate code vector and at least one ordered parameter vector may also be configured to determine a distance between at least one potential code vector and at least one ordered parameter vector based on the first and second distance components, as well as based on the values of the scale factors.

A lattice vector quantizer configured to determine at least one leading class associated with a potential codevector that generates the smallest associated distance may also be configured to determine an index of the potential codevector that generates the smallest associated distance and to determine a scale factor value of the potential codevector that generates the smallest associated distance by applying scaling factors only to the potential codevector that generates the smallest associated distance.

A trellis vector quantizer configured to transpose at least one leading class to form an output code vector derived from the trellis quantization may also be configured to sort at least one leading class by reordering at least one of the vector's absolute tuples.

A computer software product that enables the device to perform the described method.

An electronic device may include the described device.

The chipset may include the device described above.

BRIEF DESCRIPTION OF THE DRAWINGS

For a better understanding of the essence of the present invention, in the following examples, reference is made to the accompanying drawings, in which:

Figure 1 schematically shows an electronic device in which some embodiments of the present invention are used;

in fig. 2 schematically shows an audio codec system in accordance with some embodiments of the present invention;

in fig. 3 schematically shows the encoder shown in FIG. 2, in accordance with some embodiments of the present invention;

in fig. 4 shows the operation of the audio encoder shown in FIG. 3, in accordance with some embodiments of the present invention;

in fig. 5 schematically shows the lattice vector quantizer shown in FIG. 3, in accordance with some embodiments of the present invention; and

in fig. 6 shows the operation of the lattice vector quantizer shown in FIG. 5 in accordance with some embodiments of the present invention.

DESCRIPTION OF SOME EMBODIMENTS OF THE PRESENT INVENTION

Possible stereo and multi-channel speech and audio codecs are described in more detail below, including layered or scalable variable rate audio and speech codecs.

With current transform coding approaches, there may be a problem that the use of efficient gratings in compression can significantly improve quantization. However, such improvements come at the expense of a significant complication of the codec.

As discussed in detail in this description of the concept of implementation of the present invention, an approach is proposed that can significantly reduce the complexity of coding by estimating quantization distortion in the transposed vector space.

In this regard, first with reference to FIG. 1 is a block diagram of a typical electronic device or device 10 that may include a codec in accordance with an embodiment of the present invention.

Device 10 may, for example, be a mobile terminal or user equipment of a wireless communication system. In other embodiments, the device 10 may be an audio-video equipment such as a video camera, a television receiver, an audio recording or playback device, such as an mp3 file recorder/player, a multimedia file recorder (also known as a file recorder/player mp4 format) or any computer suitable for processing audio signals.

The electronic device or device 10 in some embodiments includes a microphone 11, which is connected via an analog-to-digital converter (ADC, Analogue-To-Digital Converter) 14 to the processor 21. The processor 21 is further via a digital-to-analog converter (DAC, Digital-To- Analogue Converter) 32 is connected to speaker 33. Processor 21 is also connected to transceiver (RX/TX) 13, user interface (UI, User Interface) 15 and memory 22.

The processor 21 in some embodiments of the invention may be configured to execute various program codes. Implemented software codes in some embodiments of the invention, as described below, may contain codes for encoding or decoding an audio signal. Implemented program codes 23 may in some embodiments be stored in, for example, memory 22 from which they are read by processor 21 as needed. Memory 22 may also include a partition 24 for storing data, such as data encoded in accordance with the present application.

The encoding and decoding code in embodiments of the present invention may be implemented at least in part in hardware and/or firmware.

The user interface (UI) 15 allows the user to enter commands into the electronic device 10, for example, using a keyboard, and/or receive information from the electronic device 10, for example, on a display. In some embodiments, the input and output functions of the user interface may be performed by a touch screen. Device 10 in some embodiments includes a transceiver (RX/TX) 13 that allows communication with another device, such as through a wireless communication network.

The transceiver 13 may communicate with other devices using any suitable known communication protocol, for example, in some embodiments, the transceiver 13 or transceiver may use a suitable Universal Mobile Telecommunications System (UMTS) protocol, a wireless local area network (WLAN) , Wireless Local Area Network), such as IEEE 802.X, a suitable short distance RF communication protocol such as Bluetooth or Infrared Data Communication Pathway (IRDA).

It should be understood that the structure of the device 10 can be added to or changed in various ways.

A user of device 10, for example, may use microphone 11 to input speech or other audio signals to be transmitted to some other device or stored in data section 24 of memory 22. In some embodiments, an appropriate application may be activated by the user through user interface 15 to do this. Such an application in these embodiments may be started by the processor 21 and cause the processor 21 to execute the code to be encoded and stored in the memory 22. Although, in the following examples, the microphone 11 is configured to generate audio signals for input to the device, it should be appreciated that input audio signals may be received from any suitable input device such as memory 22, in particular from the data partition 24 stored in memory 22, and in some embodiments, the input audio signal or at least one audio signal may be received via transceiver 13. For example, transceiver 13 may be configured to receive audio signals generated by microphones external to device 10, such as from Bluetooth equipment connected to the device via transceiver 13.

Analog-to-digital converter 14 in some embodiments of the invention converts the input analog audio signal into a digital audio signal and provides this signal to the processor 21. In some embodiments of the invention, the microphone 11 may include an integrated microphone and an ADC functional block and provide digital audio signals for processing directly to the processor.

The processor 21 in such embodiments then processes the digital audio signal in the same manner as described with reference to the system shown in FIG. 2, and in particular with reference to the encoder shown in FIG. 3 and the encoder elements shown in FIG. 5.

The resulting bit stream may, in some embodiments, be provided to transceiver 13 for transmission to another device. Alternatively, the encoded audio signal data in some embodiments of the invention may be stored in the data section 24 of the memory 22, for example, for later transmission or subsequent presentation by the same device 10.

The device 10 in some embodiments of the invention can also through the transceiver 13 receive from another device a bit stream containing appropriately encoded data. In this example, processor 21 may execute decoding software code stored in memory 22. Processor 21, in such embodiments, decodes received data and provides the decoded data to digital-to-analog converter 32. D/A converter 32 converts the decoded digital data into analog audio. signal and in some embodiments of the invention may output an analog audio signal through the speakers 33. In some embodiments of the invention, the execution of the decoding software code may also be activated by an application called by the user through the user interface 15.

In addition, the received encoded data, in some embodiments, may not be immediately output through the speakers 33, but stored in the data section 24 of the memory 22, for example, for later decoding and playback, or for decoding and forwarding to another device.

It will be appreciated that the schematic structures described with reference to FIGS. 3 and 5 and the method steps described with reference to FIG. 4 and 6 represent only a part of the operations performed by the audio codec, and specifically those structures and operations that correspond to the audio coding device or method shown as an exemplary implementation in the device shown in FIG. one.

The main operations performed by audio codecs and implemented according to embodiments of the present invention are shown in FIG. 2. Basic audio encoding/decoding systems comprise both an encoder and a decoder, as shown schematically in FIG. 2. However, it should be appreciated that, according to some embodiments of the invention, either only the encoder, or only the decoder, or both the encoder and the decoder may be implemented in the system. Shown in FIG. 2, system 102 includes an encoder 104, a storage or media channel 106, and a decoder 108. It should be appreciated that, as described above, in some embodiments of the invention, the system may include or be implemented with a single encoder 104, or as an encoder 104 and decoder 108.

Encoder 104 compresses input audio signal 110 and generates bitstream 112, which in some embodiments may be stored or transmitted over media channel 106. Encoder 104 in some embodiments may comprise a multi-channel encoder that encodes two or more audio signals.

Bitstream 112 may be received at decoder 108. Decoder 108 decompresses bitstream 112 and generates audio output 114. Decoder 108 may include a transform decoder that performs a portion of all decoding operations. The decoder 108 may also include a multi-channel decoder that decodes two or more audio signals. The bit rate 112 and the quality of the output audio signal 114 relative to the input signal 110 are the main characteristics that determine the performance of the encoding system 102.

In FIG. 3 schematically shows an encoder 104 in accordance with some embodiments of the present invention.

In FIG. 4 shows the operation of an encoder 104 in accordance with some embodiments of the present invention.

The described concept of the embodiments of the present invention is to define coding and apply it to audio signals to realize efficient, high quality, low bit rate coding in real time. In this regard, in FIG. 3 shows an example of an encoder 104 in accordance with some embodiments of the present invention. In addition, with reference to FIG. 4 describes the operation of the encoder 104 in more detail. In the following examples, the encoder is configured to generate frequency domain parameters representing an audio signal and encode the generated frequency domain parameters using a suitable vector trellis quantization method, however, it should be understood that in some embodiments, the described parameters, used in trellis quantization may be any suitable parameters defining or representing audio or other types of signals (eg, image or video).

The encoder 104 in some embodiments comprises a framing unit 201 or a suitable means for separating the audio signal. The framing unit 201 is configured to receive audio signals (for example, a mono signal, a stereo signal of the left and right channels, or any representation of a multi-channel audio signal), to input an audio signal, and to divide or segment the audio signal data into sections or frames suitable for conversion. frequency or other domain. Framing unit 201 in some embodiments of the invention may also be configured to window these frames or sections of audio signal data in accordance with any windowing function. For example, the framing unit 201 may in some embodiments be configured to generate 20 ms frames that overlap each preceding and following frame by 10 ms.

The audio framing operation is performed in step 501 shown in FIG. 4.

In some embodiments, audio frames may be passed to parameter determiner 203 .

In some embodiments, the encoder comprises a parameter determiner 203 including suitable means for determining at least one parameter representing the input audio signal(s) or frames of the input audio signal. In the following examples, the parameter is a Line Spectral Frequency (LSF) parameter, however, it should be appreciated that in some embodiments of the invention, any suitable parameter may be determined.

For example, in some embodiments of the invention, the parameter determiner comprises a transform 203 or a suitable transform means. Converter 203, in some embodiments, is configured to generate representations of the frequency domain (or other suitable domain) parameters of these audio signals. These frequency domain parameter representations may, in some embodiments, be passed to the parameter encoder 205 .

In some embodiments of the invention, the converter 203 may be configured to perform any suitable time-to-frequency transformation on the audio signal data. For example, the transformation from the time domain to the frequency domain can be a Discrete Fourier Transform (DFT, Discrete Fourier Transform), Fast Fourier Transform (FFT, Fast Fourier Transform), Modified Discrete Cosine Transform (MDCT, Modified Discrete Cosine Transform). The following examples use the Fast Fourier Transform (FFT).

Furthermore, the transducer may also be configured to generate separate bandwidth domain parameter representations (subband parameter representations) of the audio signal data of each input channel. These ranges can be placed in any convenient way. For example, these ranges can be arranged linearly or distributed according to perception or psychoacoustics. The generated parameters may be any suitable parameters.

The operation of determining or generating parameter representations is performed at step 503 shown in FIG. 4.

In some embodiments, representations, such as LSF parameters, are passed to parameter encoder 205 .

In some embodiments, the encoder 104 may include a parameter encoder 205 . Parameter encoder 205 may be configured to receive parameter representations of the input audio signal, such as certain LSF parameters. Encoder 205 in some embodiments of the invention can also be configured to use each of the LSF parameter values as a subvector and combine each subvector to create a vector for input to the vector quantizer. In other words, the device may comprise a vector generator configured to generate a first vector of parameters (or tuples of a first vector representing parameters) defining at least one audio signal.

The output of the vector quantizer in some embodiments of the invention is an encoder, and thus the outputs of the vector quantizer are "encoded" or encoded representations of audio signal parameters.

The parameter encoding or vector quantization operation is performed in step 505 shown in FIG. 4.

In some embodiments, the parameter encoder 205 includes a vector generator 451 . The vector generator 451 is configured to receive the LSF parameters and to generate an N-dimensional vector from these values.

The operation of generating vectors from the input parameters is performed in an additional step 551 shown in FIG. 4.

The generated vectors may, in some embodiments, be passed to the trellis vector quantizer 453.

In some embodiments, the parameter encoder 205 includes a trellis vector quantizer 453. The trellis vector quantizer 453 takes an input vector generated from the LSF parameters and generates a nearest adjacent or NN output (nearest neighbor) that appears within a certain lattice and thus , can be decoded using a similar trellis in the decoder.

The vector trellis quantization operation is performed in an additional step 553 shown in FIG. 4.

The encoded signal may be an output signal.

The operation of outputting the encoded signal is performed in step 507 shown in FIG. 4. This could be, for example, the trellis-based vector derivation operation performed in additional step 557 shown in FIG. 4.

In FIG. 5 shows an example of a trellis vector quantizer 453 in accordance with some embodiments of the present invention. The trellis vector quantizer 453 may in some embodiments be defined by the corresponding computer program code 23 that is stored in the memory 22 of the storage medium.

Before introducing the concepts and embodiments of the present invention, the standard vector trellis quantization should first be discussed. In some trellis quantizers, an initial generation or determination of a set of candidate basis codevectors is performed, with each determined basis codevector in that set of candidate basis codevectors being associated with a candidate basis codevector of another set of candidate basis codevectors.

Each set of candidate basis code vectors contains at least one basis code vector. Since each basis codevector set is associated with at least one scale factor representative of the plurality of scale factor representatives, a code vector may be determined based on a basis code vector of the set of candidate basis code vectors and a scale factor representative of at least one scale factor representative associated with a set of potential basis code vectors. In other words, a code vector may be represented based on a basis code vector scaled using an appropriate scale factor representative. For example, a scale factor representative may be a scale factor value, wherein the code vector may be determined by multiplying the basis code vector and the corresponding scale factor value. In addition, in some embodiments of the invention, the codebook is generated by applying a (signed) permutation of the basis vector.

For example, at least one set of basis code vectors is associated with at least two scale factor representatives.

Accordingly, by way of example, the codebook may comprise a codevector set containing codevectors based on a plurality of basis codevector sets and a corresponding at least one scaling factor value associated with a corresponding basis codevector set of the plurality of basis codevectors. This codevector set for each basis codevector of each basis codevector set and each at least one scale factor representative associated with the corresponding basis code vector set may comprise a code vector based on the corresponding basis code vector scaled by the corresponding scale factor representative .

For example, said sets of basis code vectors may be lead classes, each containing a different lead vector and permutations of said lead vector. Thus, said lead vector and permutations of said lead vector may represent the basis code vectors of the corresponding set of basis code vectors.

The set of basis code vector sets may represent a subset of the second set of basis code vector sets. For example, assuming that each set of basis code vectors represents a leading class, the set of leading classes may represent a subset of the second set of leading classes. Thus, the set of leading classes can be considered as a truncated set of leading classes with respect to the second set of leading classes.

For example, a corresponding candidate basis code vector may be determined by determining a basis code vector based on at least one basis code vector from the corresponding set of basis code vectors closest to the input vector to be encoded. Any kind of suitable criterion may be used to find the closest basis code vector to the input vector to be encoded.

For example, a potential basis code vector may be determined based on the closest basis code vector to an absolute value-calculated input vector and based on sign information of the input vector values, which information may include the sign of the corresponding position of the corresponding values in the input vector and be used to assigning signs to the values of a certain potential basis code vector. In addition, as an example, the basis code vector closest to the absolute value-calculated input vector containing the absolute values corresponding to the values of the input vector can be determined, with the potential basis code vector representing the determined closest basis code vector, and the signs of the values of the potential basis code vector vectors correspond to the signs of the values of the input vector at the same position in the vector, such a relationship can hold if the basis codevector parity of the basis codevector set is 0. According to another example, if the basis codevector parity of the basis codevector set is -1 , the signs of the values of the potential basis code vector can be assigned according to the signs of the values of the input vector, respectively, at the same position of the vector, and if there is an even number of negative components, the smallest non-zero absolute value in the potential basis sleeping code vector can change its sign. As another example, if the basis codevector parity of the basis codevector set is +1, the signs of the values of the potential basis codevector may be assigned according to the signs of the values of the input vector, respectively, at the same position of the vector, and if there is an odd number of negative components, the smallest non-zero absolute value in the potential basis code vector may change its sign.

The codevector for encoding the input vector is then standardly defined based on a set of defined candidate codevectors, where said set of defined candidate codevectors defines a codevector subset containing, for each defined candidate basis codevector and each scale factor representative associated with the basis code set vectors of the corresponding basis code vector, a code vector based on the corresponding candidate basis code vector, scaled by the corresponding scale factor representative.

Accordingly, a search for a code vector to encode an input vector is performed on a subset of code vectors given by the determined candidate code vectors and the corresponding at least one scale factor representative associated with the set of basis code vectors of the corresponding determined candidate code vector. Since this codevector subset may represent a subset of codevectors associated with a codebook, the number of codevectors in this codevector subset may be less than the number of codevectors in the codevector set.

For example, each scalefactor representative of the plurality of scalefactor representatives may be associated with at least one codevector set, wherein each codevector set of the at least one codevector set associated with the corresponding scalefactor representative is associated with a basis codevector set. vectors from a plurality of basis code vector sets such that each code vector set of said at least one code vector set associated with a respective scale factor representative contains code vectors obtained by scaling the basis vectors of the corresponding associated basis vector set using the respective scale factor representative .

Accordingly, the codevectors of at least one set of basis codevectors associated with a corresponding scale factor representative of the plurality of scale factor representatives may be determined based on the scaling of the basis code vectors of each set of basis code vectors associated with the scale factor representative using that scale factor representative. coefficient.

For example, if said basis codevector sets represent the leading classes, at least one basis codevector set associated with the corresponding scale factor representative can be considered as a union of the leading classes. It should be taken into account that typically the union of the leading classes does not depend on the scale factor. Thus, a codebook may comprise at least one leading class union, with each leading class union associated with at least one scale factor representative and at least one basis codevector set of the set of basis codevectors. For example, at least one scale factor representative may represent a plurality of scale factor representatives, which may include at least two scale factor representatives.

{0, 1, …, X-1} представляет набор базисных кодовых векторов из множества наборов базисных кодовых векторов, при этом X представляет количество наборов в множестве наборов базисных кодовых векторов. Каждый набор базисных кодовых векторов связан по меньшей мере с одним базисным кодовым вектором b_x,y или содержит этот вектор, где B_x представляет количество базисных кодовых векторов соответствующего набора базисных кодовых векторов b_x, то есть выполняется соотношение y

{0, 1, …, B_x - 1}. Например, количество B_x базисных кодовых векторов в наборе базисных кодовых векторов может различаться для различных наборов базисных кодовых векторов, и/или оно может быть одинаковым по меньшей мере для двух наборов базисных кодовых векторов.So, for example, b _x , for x

{0, 1, ..., X-1} represents a set of basis code vectors from the set of basis code vector sets, where X represents the number of sets in the set of basis code vector sets. Each set of basis code vectors is associated with or contains at least one basis code vector b _x,y , where B _x represents the number of basis code vectors of the corresponding set of basis code vectors b _x , i.e. y

{0, 1, …, B _x - 1}. For example, the number B _x of basis codevectors in a basis codevector set may be different for different basis codevector sets and/or it may be the same for at least two basis codevector sets.

In other words, there is only one leading vector. Together with all signed permutations of the leading vector, this set forms the leading class of the leading vector (or, as noted in this description, basis code vectors). If you connect several leading classes, a union of leading classes is formed. One or more scale factors can then be attached to this pool(s).

{0, 1, …, S - 1}.Thus, for example, it is possible to determine a code vector c _x , _z , _y based on a basis code vector b _x,y and a scale factor representative s _z , where the index z represents the index of the corresponding scale factor representative in the set of scale factor representatives s _o … s _{S -1} , that is, the relation z

{0, 1, ..., S - 1}.

{0, 1, … n-1}, и n представляет длину соответствующего базисного кодового вектора b_x,y, и если вычисленный по абсолютному значению входной вектор используется для определения потенциального кодового вектора соответствующего набора базисных кодовых векторов, знак каждого значения b_x,_y,_t в (t+1)-ой позиции определенного ближайшего базисного кодового вектора b_x,_y может назначаться на основе знака соответствующего значения i_t в (t+1)-ой позиции входного вектора i перед определением кодового вектора c_x,_z,_y на основе базисного кодового вектора b_x,_y и представителя масштабного коэффициента s_z.For example, if the values b _x,y,t of the basis code vectors b _x,y = [b _x,y.0 , b _x,y,1 , … b _x,y,n-1 ] represent absolute values , where t

{0, 1, ... n-1}, and n represents the length of the corresponding basis code vector b _x,y , and if the absolute-valued input vector is used to determine the potential code vector of the corresponding set of basis code vectors, the sign of each value of b _x , _y , _t at the (t+1)th position of the determined closest basis code vector b _x , _y may be assigned based on the sign of the corresponding value i _t at the (t+1)th position of the input vector i before determining the codevector c _x , _z , _y based on the basis code vector b _x , _y and scale factor representative s _z .

For example, if i = [i ₀ , i ₁ , …, i _n-1 ] represents an input vector, then the absolute value of the input vector can be represented as follows: [|i ₀ |, |i ₁ |, …, | i _n-1 |]. For example, the sign of each b _x,y,t value at the (t+1)th position of the specified nearest basis b _x,y code vector can be the same as the sign of the corresponding i _t value at the (t+1)th position of the input vector, respectively , and this condition can be met if the parity of basis code vectors b _x,y of the set of basis code vectors b _x is equal to 0. In another example, if the parity of basis code vectors b _x , _y of the set of basis code vectors b _x is equal to -1, then the signs of the values b _x,y,t of the candidate basis code vectors may be assigned according to the signs of the values of the input vector at the same position in the vector, respectively, and if there is a non-odd number of negative components, the value of b _x , _y , _t in the potential basis code the vector with the smallest non-zero absolute value can change its sign. In another example, if the parity of the basis code vectors b _x , _y of the set of basis code vectors b _x is +1, the signs of the values of the potential basis code vector b _x,y,t can be assigned according to the signs of the values of the input vector respectively at the same position of the vector , and if there is an odd number of negative components, the value b _x,y,t in the potential basis codevector with the smallest non-zero absolute value may change sign.

In one example, the code vector c _x,z , _y can be defined as follows: c _x , _z , _y = [b _x,y,0 ⋅ s _z , b _x,y,1 ⋅ s _z , …, b _{x, y,n-1} s _z ].

{0, 1, …, S-1}, связан по меньшей мере с одним набором базисных кодовых векторов. Например, согласно одному из примеров этот соответствующий по меньшей мере один набор базисных кодовых векторов может быть представлен набором базисных кодовых векторов b_x при x

{0, 1, … n_z - 1}, при этом n_z может представлять количество наборов базисных кодовых векторов, связанных с соответствующим представителем масштабного коэффициента s_z, где 0 < n_z < X. На основе этой связи между соответствующим представителем масштабного коэффициента s_z и по меньшей мере одним ассоциированным набором базисных кодовых векторов b_x при x

{0, 1, … n_{z - 1}} может быть определен по меньшей мере один ассоциированный набор базисных кодовых векторов c_x,z,y, где x

{0, 1, …, n_z - 1}, y

{0, 1, …, B_x - 1} и z

{0, 1, …, S - 1}.Each representative of the scale factor s_z, where z

{0, 1, ..., S-1} is associated with at least one set of basis code vectors. For example, in one example, this corresponding at least one basis code vector set may be represented by the basis code vector set b_x at x

{0, 1, … n_z - 1}, while n_z may represent the number of basis codevector sets associated with the corresponding scale factor representative s_z, where 0 < n_z < X. Based on this relationship between the respective scale factor representative s_z and at least one associated set of basis code vectors b_x at x

{0, 1, … n_{z - 1}} at least one associated set of basis code vectors c_x,z,y, where x

{0, 1, …, n_z - 1}, y

{0, 1, …, B_x - 1} and z

{0, 1, ..., S - 1}.

Thus, for example, the structure of the aforementioned codebook may be defined by a plurality of scale factor representatives s _z , a plurality of basis code vector sets b _x , and an association between each scale factor representative with at least one associated basis code vector set.

Since at least one basis codevector set, such as at least the b ₀ basis codevector set, is associated with at least two scalefactor representatives, the same basis codevector set may be used to generate the codevectors of at least one codevector set associated with the first scale factor representative, and for generating code vectors of at least one set of code vectors associated with at least one other scale factor representative.

For each basis code vector set of the plurality of basis code vector sets, it is possible in other ways to determine a potential basis code vector for encoding the input vector.

For example, the determination of a codevector to encode an input vector from a subset of codevectors is based on the determination of a distortion metric, or a distance, or an error value.

In such examples, a scale factor representation is selected from a plurality of scale factor representations.

In addition, a certain potential basis code vector is selected from the set of basis code vectors associated with the selected scale factor representation.

A codevector may then be determined based on the selected candidate basis codevector and the selected scale factor representation, and such codevector determination may be performed according to the method presented herein.

In some examples, based on a certain code vector and input vector, a distortion metric is determined. For example, said distortion metric may be based on any kind of suitable distance between a particular code vector and an input vector. For example, the Hamming distance or the Euclidean distance, or any other distance, may be used. For example, the definition of a code vector may be omitted, and the distortion metric may be calculated by specifically considering the corresponding code vector associated with the selected scalefactor representation and the set of basis codevectors associated with the selected scalefactor representation.

For example, if c _x , _z , _y = [c _x , _z , _y , _o , c _x , _z , ₁ , ..., c _x , _z , _n-1 ] represents the code vector determined in step 430 and i = [i ₀ , i ₁ , …, i _n-1 ] represents the input vector, the distance d can be calculated using the following formula:

This distance d, represented by the above formula, can be replaced by a distance d', calculated according to the following formula:

In another example, if the distortion factor is determined based on a weighting function, the distance d represented by the above formula can be modified as follows:

where w _k represents the weight coefficients of the weight function.

Accordingly, the distance d' represented by the above formula can be weighted by a weight function as follows:

For example, the distortion metric d or d' or d _w or d _w ^' may be stored if it is the first determined distortion metric, or it may be compared with the stored distortion metric, with the stored distortion metric replaced if the new determined distortion metric is better. saved. In addition, a code vector associated with the stored distortion metric or an identifier of that code vector may be stored.

Then, for example, a check may be performed to see if there are any subsequent sets of basis code vectors associated with the selected scale factor representation. If so, then a certain potential basis codevector is selected from this next set of basis codevectors associated with the selected scalefactor representation. If not, a check is made with the next scale factor representation from the set of scale factor representations.

If the next scale factor representation exists from the set of scale factor representations, then the next scale factor representation is selected, otherwise, the code vector associated with the best distance measure may be selected to encode the input vector.

For example, the sets of basis code vectors may be lead classes, each containing a different lead vector and permutations of said lead vector. Thus, the lead vector and permutations of said lead vector may represent the basis code vectors of the corresponding set of basis code vectors. For example, a leading vector is an n-dimensional vector (where n is an integer) whose (positive) components are ordered (eg, in descending order). The leading class corresponding to the leading vector includes the leading vector and all vectors obtained by all signed permutations of the leading vector (with some possible restrictions).

The union of the leading classes may be defined by a set of basis code vectors associated with the same scale factor representation of the set of scale factor representations and the corresponding scale factor representation. For example, a union of leading classes may be associated with a set of codevectors obtained by scaling the basis codevectors of the associated set of basis codevectors using a scale factor representation.

Such a union of leading classes can be considered as a truncation. Thus, if there are n scalefactor representations in the set of scalefactor representations, then n leader class unions can be defined, each defined by a corresponding scalefactor representation and a set of basis codevectors associated with the corresponding scalefactor representation.

Accordingly, a plurality of scaling factor representations and a plurality of basis code vector sets may define a union of leading classes and thus define a codebook, whereby, for example, each union of leading classes may be considered a union of scaled leader classes.

The codebooks used in these speech and audio codecs may, for example, be based on trellis structures, as described in the reference document by A. Vasilache, B. Dumitrescu and I. Tabus "Multiple-scale leader-lattice VQ with application to LSF quantization" (leading multi-scale VQ grating applied to LSF quantization), published in Signal Processing 2002, Vol. links. For example, a D10+ lattice may be considered for quantization, but any other well-suited lattice quantization schemes may also be considered.

For example, the sets of basis code vectors are leading classes, with each leading class containing different leading vectors and permutations of the specified leading vector, and each leading class representing an n-dimensional vector containing n absolute values, sorted in descending or ascending order.

The leading vector I of the corresponding set of basis code vectors b _x can be represented as follows: I = [I ₀ , I ₁ , …, I _n-1 ], where l ₀ , I ₁ , …, I _n-1 are absolute values . In the case of descending order, I ₀ represents the 1st largest value, I ₁ represents the 2nd largest value, and I _n-1 represents the nth largest value. In the case of ascending order, I ₀ represents the 1st smallest value, I ₁ represents the 2nd smallest value, I _n-1 represents the nth smallest value.

The value I _{k-1 of} the corresponding leading vector, which represents the value at the kth position in the corresponding leading vector, may be assigned to the position in the candidate basis codevector corresponding to the position of the kth largest absolute value (in the case of a descending leading vector) or position the k-th smallest absolute value (in the case of an ascending leading vector) in the input vector. For example, this position may be referred to as position m. For example, a potential basis code vector can be represented as p = [p ₀ , p ₁ , ..., p _n-1 ].

In one possible example, the input vector can be represented as follows: i = [-2.4, 5.0, -1.3, 0.2], while the corresponding vector calculated from the absolute value can be represented as follows: ia = [2.4, 5.0, 1.3, 0.2].

In the case of descending ordering of the lead vector, the value at position "k" of the lead vector, i.e., the value I _k-1 , is assigned to the position of the potential basis code vector that corresponds to the position of the k-th largest absolute value in the input vector. For example, starting from the first position represented by count k=1, the position of the 1st largest absolute value in the input vector is position m=2, because the value 5.0 is the 1st largest value of the computed absolute value of the input vector and it is located at position m =2, i.e. ia ₁ . Accordingly, the value of I ₀ is assigned to position m=2 in the potential basis code vector, that is, p ₁ = I ₀ can be satisfied.

In addition, the sign (+ or -) of the assigned value in the potential basis code vector p _m-1 is set according to the sign of the value in the input vector associated with the k-th largest absolute value. Accordingly, the equality
p _m-1 = I _k-1 ⋅ sign(i _m-1 ).

Thus, according to one example of an input vector, the relations i = [-2.4, 5.0, -1.3, 0.2], p = I ₀ can hold, since the value of i ₁ = 5.0 is positive.

The "to" position counter may be incremented, and a check may be made to see if another value exists in the leading vector, ie, if the inequality k < n is satisfied.

If the value exists, the method continues, and in one example, for position k = 2, the value 2.4 at position m=1 represents the 2nd largest (kth largest) absolute value in the input vector. Thus, the relation P ₀ = l ₁ ⋅ sign(i ₀ ) = -l ₁ can hold for sign-sensitive assignment of I ₁ because the value of I ₀ = -2.4 in the input vector is negative.

Thus, in one example, the loop can be iterated by moving to subsequent positions of the leading vector as follows:

k = 3 → m = 3 → P ₂ = I ₂ ⋅ sign(i ₂ ) = -I _{2 ;} and

k = 4 → m = 4 → P ₃ = I ₃ ⋅ sign(i ₃ ) = +I ₃

Thus, the corresponding potential basis code vector obtained in the example way, as a result, may look like this: p = [-I ₁ , I ₀ , -I ₂ , I ₃ ] if the corresponding leading vector I is ordered in descending order of values.

If the corresponding leading vector I is ordered in ascending order, then the above method can be performed using the value m representing the position of the k-th smallest value in the absolute value-calculated input vector, whereby the following relationship can be satisfied: p _m-1 = I _{k -1} ⋅ sign(i _m-1 ).

The resulting potential basis code vector p is associated with the corresponding set of basis code vectors b_x, while I represents the leading vector of this corresponding set of basis code vectors. For example, in the above example of a codevector determination process based on the basis codevector b_x,_y,_t and the representative of the scale factor s_z, the potential code vector p represents the nearest basis code vector b_x,_y set of basis code vectors b_xwith respect to the input vector, whereby the input vector calculated from the absolute value is used to determine the potential code vector of the corresponding set of basis code vectors, and the sign of each value b_x,_y,_k-1 at the kth position of the determined nearest basis code vector b_x,_y assigned using the sign of the corresponding i value_k at the kth position of the input vector i, where 0 <k ≤ n.

Thus, this nearest basis code vector b _x , _y representing a potential code vector p can be used to determine a code vector c _x , _z , _y based on the nearest basis code vector b _x , _y and the corresponding scale factor representative s _z , as has been described above.

For each truncation, different scale factor representatives are assigned (for example, by training), for example:

floatscale[] = {0.8, 1.2, 2.7};

Accordingly, for example, a first set of codevectors from the set of codebook codevectors is defined by a first truncation scaled by a first scaling factor representation of 0.8, a second set of codevectors from a set of codebook codevectors is defined by a second truncation scaled by a second scaling factor representation of 1.2, and a third the set of code vectors from the set of code vectors of the code book is determined by the third truncation, scaled by the third representation of the scale factor 2.7, as a result, the code book has a lattice structure with many scale factors.

As an example, searching in a multi-scale-factor lattice structure can be considered in two phases: in the first phase, the potential code vector for each leading class, i.e., for each set of basis code vectors, can be calculated, and in the second phase, distortion for only potential code vectors.

For example, an absolute value calculation function can be applied to an input vector i such that the absolute input vector ia contains the absolute values of vector i, and then the absolute input vector can be sorted in descending (or alternatively ascending) order.

For example, an index representation may include representatives indicating the indices of each input vector i, whose absolute values are in descending (or ascending) order. For example, the specified index representation may be an array 'indx' of integers.

For example, if the input vector is represented as [-2.4 5.0 -1.3 0.2], the absolute value vector is represented as [2.4 5.0 1.3 0.2], and the 'indx' array is represented as [1 0 2 3]. Since the leading vectors can be ordered in descending order, during the execution of the algorithm for finding the nearest adjacent vector, the first value of the leading vector can be assigned to the position corresponding to the component with the largest absolute value of the input vector, and so on.

In the following non-limiting examples, the variable 'idx_lead_max' is intended to store the maximum number of lead classes in all truncations that X can match. In this example, there may be 9. Accordingly, 9 sets of basis codevectors are defined by 9 lead classes , with the nth leading class defined by the reference &pl[n-1].

For example, the 'sign' array can store the signs of the components of the input vector.

/* First part of search: compute all potential codevectors */

An outer loop, defined by a counter u, can be considered to associate each value of u with the corresponding leading vector. Thus, according to the counter u, a corresponding set of basis code vectors is selected using an outer loop, since each leading vector corresponds to a different set of basis code vectors from the set of basis code vectors.

The inner loop, defined by the integer value j, can be considered to determine the potential basis codevector associated with the selected basis codevector set, with j_crt indicating the position of the (j+1)th largest absolute value in the input vector.

Thus, various potential basis codevectors cv_pot are determined by this example of the first part of the search.

The second part of the search may be used to determine a code vector to encode the input vector based on a subset of the code vectors.

/* Second part of the search */

The outer loop may be defined by a counter I, which is introduced to select one scale factor representation scale[I] from a plurality of scale factor representations.

The variable LATTICE_DIM specifies the length of the codevectors, which may correspond to the length of the input vector to be encoded.

Then the values ws1[k] and ws2[k] are calculated for each "k" from the series (0, ..., LATTICE_DIM), which can be considered as part of the distortion index (X3), which does not depend on the potential basis code vector. The value w[k] represents the value of the weight function for each k.

The code example above also has an inner loop implemented with counter j: "for(j=0;j<no_leaders[I];j++)", in which no_leaders[I] defines the set of leader vectors associated with the selected scale factor representative scale[I], i.e. no_leaders[I] may correspond to a value of n _z representing the number of basis code vector sets associated with the corresponding scale factor representative scale[I], and thus this loop is iterated with each set of leader vectors, associated with the selected scale factor representative scale[I], with one potential basis code vector cv_pot defined for the leading vector of this set of leading vectors. Thus, for example, this loop iteratively selects each potential basis codevector cv_pot from the set of basis codevectors associated with the selected scalefactor representation, wherein cv_pot[j] may represent the corresponding jth basis codevector of this set of basis codevectors.

For each of these basis codevectors and the selected scalefactor representative, a corresponding distortion metric for the codevector associated with the corresponding basis codevector and the selected scalefactor representative may be determined, for example, as follows:

.

.

To represent the best distortion metric, the distortion metric with the smallest value is determined, and the codevector associated with the codevector with that distortion metric can be used to encode the input vector. For example, this codevector may be determined by the best scale factor representation and the best candidate basis codevector from the set of candidate basis codevectors.

The embodiments of the present invention described in this description reduce the complexity of the vector quantization process, not by computing an array of potential codevectors cv_pot, but by applying an absolute-sorted version of the input vector and determining or generating a distortion calculation in an appropriate transposed space.

In some embodiments, the trellis vector quantizer comprises an input vector sorter 402 . An input vector sorter 402 or a suitable input vector sorter may be configured to receive an input vector.

The input vector receive operation is performed at step 501 shown in FIG. 6.

The lattice vector quantizer and input vector sorter 402 are configured to sort the input vector in descending order of absolute values (note that in some embodiments, sorting may be performed in ascending order of absolute values, with appropriate modifications to the operations below).

Thus, if, for example, the input vector is represented as follows:

I = [-2.4 5.0 -1.3 0.2],

vector computed by absolute value looks like this:

absi = [2.4 5.0 1.3 0.2],

sorted vector computed by absolute value, in this case is defined as follows:

cv_pot1 = [5.0 2.4 1.3 0.2],

and index 'indx' of sorting permutations = [1 0 2 3].

The input vector sort operation is performed at step 503 shown in FIG. 6.

The input vector sorter may then pass the sorted vector and sort permutation information to the code vector determinant 403 .

In some embodiments, the implementation of the trellis vector quantizer 453 contains the determinant 401 of the potential code vector. A candidate codevector determiner 401 or a suitable means for determining a candidate codevector is configured to store or generate leading classes used to generate codevectors.

For example, leading classes could be defined as follows (in the value of Q1, in other words, multiplied by 2):

These leading classes, in some embodiments, may be passed to the code vector determinant 403 .

In some embodiments, the trellis vector quantizer 453 includes a code vector determinant 403 . The code vector determiner 403 or a suitable code vector determiner may, in some embodiments, accept leading classes as well as a sorted input vector and a permutation vector. The codevector determiner can then, based on these values, determine the output codevector associated with the input vector.

If the distance to be determined is a weighted Euclidean distance, then in some embodiments the weights are transposed according to the permutation vector and the intermediate product of the input vector is formed, taking into account that in some embodiments the weights are constant or the weighting operation is optional, with This uses the unweighted Euclidean distance.

An example of such a process is represented by the following code.

/* Calculate the intermediate product between the transposed weights and the sorted input vector */

The operation of transposing and applying weights to form an intermediate product based on the sorted input vector and the transposed weights is performed in step 505 shown in FIG. 6.

In some embodiments, the codevector determiner may determine the distance components sum1 and sum2 for the first value scale[0] of the scale factor.

This operation can be performed through the following steps.

First, the scale factor and the square of the scale factor values for the first scale factor scale[0] are initialized.

The operation of initializing the scale factor and the square of the scale factor values is performed at step 508 shown in FIG. 6.

Second, a leading vector is selected from the matrix of leading classes. An example matrix is shown above as the pl_crt matrix.

The drive vector selection operation is performed at step 507 shown in FIG. 6.

Thirdly, intermediate values sum1 and sum2 of the distance are formed based on the intermediate values and the selected leading vector.

The operation of generating intermediate values based on the selected leading vector is performed at step 509 shown in FIG. 6.

Fourth, parity conditions are checked when the leading vector does not reach the ^7th position, and the value of sum1 is adjusted if the number of negative signs in the input vector does not satisfy the limit imposed on the leading class parity.

The operation of checking for parity conditions when the leading vector does not reach the ^7th position and adjusting the value of sum1 if the number of negative signs in the input vector does not satisfy the limit imposed on the parity of the leading class is performed in step 511 shown in FIG. 6.

Fifth, the distance or error value is determined based on the values of sum1 and sum2, and if the current distance for the leading vector is the smallest, the index of the smallest vector is indicated.

The operation of determining the distance for the leading vectors is performed at step 513 shown in FIG. 6.

The operation can then be repeated in a loop until all leading vectors have been selected.

The operation of checking if all of the leading vectors have been selected and looping back if not all of the leading vectors have been selected is performed in step 514 shown in FIG. 6.

These steps can be implemented with the code below.

Then, in some embodiments, the code vector determinant may be configured to use the values of sum1 and sum2 to determine distortion distances for other scale factors. In addition, a similar operation of checking for the "best" value of the scale factor is performed.

The operation of determining the distortion distance for other scale factors is performed at step 515 shown in FIG. 6.

The operation of determining the distortion distance for other scale factors using the values of sum1 and sum2 can be implemented by the following code example.

Furthermore, in some embodiments, the codevector determiner may be configured to calculate the resulting 'cv_out' codevector after finding the best leading class and best scaling factor.

The operation of performing an inverse transposition for code vector calculation is performed at step 517 shown in FIG. 6.

In some embodiments, the code vector calculation operation may be implemented by the following code example.

/* reverse permutation */

In some embodiments, the calculation of the variables sum1 and sum2 is performed until the number of leaders from the first truncation (no_leaders[0]) is reached, which means that the number of leaders must be ordered in descending order, as well as their corresponding scaling factors.

In such embodiments, the complexity is further reduced since it is not necessary to calculate the maximum number of leader vectors for one structure, since it is known to be in the first position.

It should be taken into account that the complexity is reduced to the greatest extent due to the fact that not all leading vectors have to be transposed, but only the best of them. The calculation is performed using positive values (both the leading vector and the input vector are represented by absolute values), which is valid as long as the components of the input and quantized vectors have the same sign.

The difference in signs is affected if there is a parity constraint (odd or even number of negative components) in the considered leading vector and this constraint is not taken into account by the input vector. In this case, the sign of the quantized value of the smallest components of the input vector is reversed. The smallest component of the input vector corresponds to the last component in the transposed space. So the first loop of calculating sum1 and sum2 looks like this: "while(l < LATTICE_DIM - 1)". Actually the non-transposed space in this case corresponds to the smallest value = indx[LATTICE_DIM - 1]. LATTICE_DIM is the size of the lattice in question.

While the above examples describe embodiments of the present invention operating in device 10, it should be appreciated that the invention, as described below, may be implemented as part of any audio (or speech) codec, including any audio (or speech) codec. signal with variable/adaptive speed. Thus, for example, embodiments of the present invention may be implemented in an audio codec that can encode audio signals transmitted over fixed or wired lines.

Thus, the user equipment may contain an audio codec similar to those described in the above embodiments of the present invention.

It should be noted that the term user equipment encompasses any suitable type of wireless communication user equipment, such as mobile phones, portable data processing devices, or portable web browsers.

In addition, the public land mobile network (PLMN) elements may also contain the audio codecs described above.

In general, various embodiments of the present invention may be implemented in hardware or dedicated circuits, software, logic circuits, or any combination of these means. For example, some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device, although the invention is not limited to these. Although various aspects of the present application may be illustrated and described in the form of block diagrams, algorithms or using some other graphical representations, it is quite obvious that the blocks, devices, systems, technologies or methods described here can be implemented (without being limited to the examples given) in in the form of hardware, software, firmware, dedicated circuits or logic circuits, general purpose hardware or controllers, or other computing devices, or some combination of these.

Embodiments of this invention may be implemented by computer software executed by a data processor of a mobile device, such as a processor unit, or by hardware, or by a combination of software and hardware. Also, in this regard, it should be noted that the various logic blocks shown in the drawings may be program steps or interrelated logic circuits, blocks and functions, or a combination of program steps and logic circuits, blocks and functions.

The memory may be of any type appropriate to the local technical environment and may be implemented using any suitable storage technology and may be, for example, semiconductor memory devices, magnetic memory devices and systems, optical memory devices and systems, read-only memory, and removable memory blocks. Data processors may be of any type suitable for the local technical environment and may, for example, include one or more general purpose computers, specialized computers, microprocessors, digital signal processors (DSP, Digital Signal Processor), specialized integrated circuits (ASIC, Specific Integrated Circuit ), logic circuits and processors based on multi-core architecture, as well as other similar devices.

Embodiments of the present invention may be practiced in various components such as integrated circuit modules. In general, the design of integrated circuits is a highly automated process. Comprehensive and efficient software tools are available to convert a logic level design into a semiconductor circuit prepared for etching and semiconductor wafer formation.

Programs produced by, for example, Synopsys, Inc., located in Mountain View, California, and Cadence Design, located in San Jose, California, automatically route conductors and place components on a semiconductor chip using well-defined design rules, as well as libraries , which store pre-recorded structural modules. Once the semiconductor circuit design is complete, the resulting design in a standardized electronic format (eg, Opus, GDSII, etc.) can be transferred to a semiconductor device manufacturing facility or fabrication module ("fab") for fabrication.

As used in this application, the term "circuit" refers to all of the following components:

only to hardware implementations of circuits (such as implementations made only in the form of analog and/or digital circuits); and

to a combination of circuitry and software (and/or firmware) such as: (i) a combination of processor(s) or (ii) a combination of processor(s)/software (including digital signal processor(s)), software software and memory that together enable a device, such as a mobile phone or a server, to perform various functions; and

(c) circuitry, such as a microprocessor(s) or a component of the microprocessor(s), that requires software or firmware to operate, even if it is not physically installed.

This definition of "circuit" applies to all uses of the term in this application, including any claims. As another example of the term "circuitry" as used in this application, one can cite an implementation of a conventional processor (or multiple processors), or a component of the processor and the software or firmware that it (or they) execute. The term "circuit" also refers, for example (including, if applicable to a particular claim), to a baseband signal processing integrated circuit or an application processor integrated circuit in a mobile phone, or a similar integrated circuit in a server, in cellular network device or other network device.

The foregoing description, by way of non-limiting generic examples, provides a complete and informative description of the exemplary embodiments of the present invention. However, a person skilled in the art in the light of the foregoing description, read in conjunction with the accompanying drawings and claims, may be obvious to various modifications and adaptations. However, any kind of such and similar modifications to the principles set forth shall remain within the scope of the present invention as defined in the appended claims.

Claims (49)

1. A method for encoding audio signals, including: generating at least one vector of parameters defining at least one audio signal; sorting the components of said at least one parameter vector according to a descending order based on the absolute values of the components of said at least one parameter vector to form at least one associated ordered parameter vector; selecting at least one potential code vector from the list of leading classes; determining the distance between at least one potential code vector and at least one ordered parameter vector by generating the first and second distance components based on the first scale factor value, wherein the first and second distance components are related to the distance between at least one potential code vector and at least one ordered parameter vector, correcting the sign of the first distance component when at least one candidate code vector fails a parity constraint test, and determining the distance between the at least one candidate code vector and the at least one ordered parameter vector by based on the first and second distance components; determining at least one leading class associated with the potential code vector that generates the smallest associated distance; and sorting at least one leading class by reverse ordering in descending order based on the absolute values of the components of said at least one parameter vector. 2. The method according to p. 1, also including: selecting at least one scale factor; while determining the distance between at least one potential code vector and at least one ordered parameter vector also includes determining the distance based on at least one scale factor; determining at least one leading class associated with the candidate code vector that generates the smallest associated distance also includes determining an output scale factor associated with said candidate code vector and the scale factor that generates the smallest associated distance; and sorting the at least one leading class to generate an output trellis-quantized codevector includes applying an output scaling factor to the trellis-quantized output codevector. 3. The method according to claim 1 or 2, characterized in that the formation of the first vector of parameters that define at least one sound signal includes: dividing at least one audio signal into time frames; and determining a vector of frequency parameters of the discrete spectrum associated with at least one of the time frames of the audio signal. 4. The method according to claim 1 or 2, characterized in that sorting the components of said at least one parameter vector according to descending order based on the absolute values of the components of said at least one parameter vector to form at least one associated ordered parameter vector comprises: defining weights to determine the weighted distance; sorting the weights based on a descending order based on the absolute values of the components of said at least one parameter vector to generate a vector of sorted weights; and applying the sorted weight vector to at least one ordered parameter vector. 5. The method according to claim. 1, characterized in that determining the distance between at least one potential code vector and at least one ordered parameter vector also includes determining the distance between at least one potential code vector and at least one ordered parameter vector based on the first and second distance components, as well as based on scale factor values. 6. The method according to claim 1 or 2, characterized in that determining at least one leading class associated with a potential code vector that generates the smallest associated distance includes: determining an index of a potential code vector that generates the smallest associated distance; and determining a scale factor value of the potential code vector that generates the smallest associated distance by applying the scale factor values only to the potential code vector that generates the smallest associated distance. 7. A device for encoding audio signals, comprising: means for generating at least one vector of parameters defining at least one audio signal; means for sorting the components of said at least one parameter vector according to a descending order based on the absolute values of the components of said at least one parameter vector to form at least one associated ordered parameter vector; means for selecting at least one potential codevector from a list of leading classes; means for determining the distance between at least one potential code vector and at least one ordered parameter vector by generating the first and second distance components based on the first scale factor value, wherein the first and second distance components are associated with the distance between at least one potential code vector and at least one ordered parameter vector, correcting the sign of the first distance component when at least one candidate code vector fails a parity constraint test, and determining the distance between at least one candidate code vector and at least one ordered vector parameters based on the first and second distance components; means for determining at least one leading class associated with the potential code vector that generates the smallest associated distance; and means for sorting at least one leading class by reverse ordering in descending order based on the absolute values of the components of said at least one parameter vector. 8. The device according to claim 7, also including: means for selecting at least one scale factor; the means for determining the distance between the at least one potential code vector and the at least one ordered parameter vector also comprises means for determining the distance based on the at least one scale factor; means for determining at least one leading class associated with the potential code vector that generates the smallest associated distance also includes means for determining an output scale factor associated with said potential code vector and the scale factor that generates the smallest associated distance; and means for sorting at least one leading class to generate an output code vector derived from the trellis quantization comprises means for applying an output scaling factor to the output code vector derived from the trellis quantization. 9. The device according to claim 7 or 8, characterized in that the means for generating the first vector of parameters defining at least one sound signal includes: means for separating at least one audio signal into time frames; and means for determining a vector of frequency parameters of the discrete spectrum associated with at least one of the time frames of the audio signal. 10. Apparatus according to claim 7 or 8, characterized in that the means for sorting the components of said at least one parameter vector according to descending order based on the absolute values of the components of said at least one parameter vector to form at least one associated ordered parameter vector contains: means for determining weights for determining the weighted distance; means for sorting the weights based on a descending order based on the absolute values of the components of said at least one parameter vector to generate a vector of sorted weights; and means for applying the sorted weight vector to the at least one ordered parameter vector. 11. The device according to claim. 7, characterized in that the means for determining the distance between at least one potential code vector and at least one ordered parameter vector also contains means for determining the distance between at least one potential code vector and at least one ordered vector of parameters based on the first and second distance components, as well as based on the scale factor values. 12. The device according to claim 7 or 8, characterized in that the means for determining at least one leading class associated with a potential code vector that generates the smallest associated distance comprises: means for determining an index of the potential code vector that generates the smallest associated distance; and means for determining a scale factor value of the potential code vector that generates the smallest associated distance by applying the scale factor values only to the potential code vector that generates the smallest associated distance. 13. A storage medium containing a computer program that, when executed, causes the device to perform the method according to any one of paragraphs. 1-6. 14. An electronic device for encoding audio signals, containing the device according to paragraphs. 7-12. 15. Chipset, including the device according to paragraphs. 7-12.

Images