37 results about "Scalar quantization" patented technology

This provides a concept that enables coding of the coefficients of a transformation block in a manner suitable for dependent quantization and, in terms of coding efficiency, can be effectively implemented by entropy coding. [Solution] In a video encoder, dependent scalar quantization is utilized in entropy coding by using different sets of probabilistic models for different sets of acceptable reconstruction levels, where the set of acceptable reconstruction levels for the transformation coefficients depends on the transmitted transformation coefficient levels that precede the current transformation coefficient in the reconstruction order.

The invention discloses a lightweight lossless audio encoding and decoding method and system, and belongs to the technical field of audio signal processing. The method comprises the following steps: converting an original audio waveform into a compact coding expression through an encoder formed by connecting a TConv unit, a convolution unit, a down-sampling unit and a Local-Transform unit in series; a single finite scalar quantizer is adopted to discretize the coded representation, and a mixed quantization strategy is introduced to generate a single and consistent discrete token stream; and finally, a high-fidelity audio waveform is reconstructed through a decoder formed by connecting a Local-Transform unit, a TEConv unit, an up-sampling unit and an audio recovery unit in series. Through unit structure optimization and joint loss function training, indexes such as SDR, PESQ, STOI, WER and the like under multiple different code rates are all superior to those of a baseline model, model parameters are only 10.31-11.29 M, collaboration of lightweight deployment and high-fidelity reconstruction is achieved, resource-constrained equipment is adapted to downstream scenes such as voice communication, ASR and the like, and good engineering practicability is achieved.

To provide a method and apparatus for video encoding or decoding.SOLUTION: When the method uses dependent scalar quantization, the selection of the quantizer depends on the decoding of the preceding transform coefficient, and the entropy decoding of the transform coefficient depends on the selection of the quantizer. In order to maintain high throughput in a hardware implementation of transform coefficient entropy coding, several decision schemes for scalar quantizers are proposed, where the selection of state transitions and context models is based only on the regular coded bins. For example, the state transition is based on the sum of the SIG, gt1 and gt2 flags, the exclusive-OR functions of the SIG, gt1 and gt2 flags, or only the gt1 or gt2 flags. When a block of transform coefficients is coded, the regular mode bins are coded first in one or more scan passes, and the remaining bypass-coded bins are grouped in another one or more scan passes.SELECTED DRAWING: Figure 11

In various embodiments, an encoder receives transform coefficients of prediction residues associated with video data. The encoder sets a flag value based on at least one of metadata or transform statistics associated with a vector of the transform coefficients. The encoder performs trellis coded quantization operations on the vector of the transform coefficients to generate a vector of quantization indices or one or more scalar quantization operations on the transform coefficients to generate a set of quantization indices based on the flag value. The encoder performs entropy coding operations on either the vector of quantization indices or a vector corresponding to the set of quantization indices to generate encoded video data.

The invention belongs to the technical field of communication, and particularly relates to a semantic communication method combining variable length coding and constellation probabilistic shaping, which comprises the following steps of: firstly, constructing a semantic encoder based on a mask auto-encoder framework, and designing a mask rate control function through channel state information and information source characteristics so as to dynamically adjust an information source mask rate; a self-adaptive variable-length coding mechanism is realized; according to the obtained mask rate, sampling a non-masked information source unit by adopting a maximum entropy criterion so as to reserve the information amount of the information source to the maximum extent; the constellation point optimal probability is shaped and approximated to discrete two-dimensional Gaussian distribution, and an end-to-end optimization target is deduced based on variational inference; semantic feature coding distribution is constrained through layer normalization operation to approximate Gaussian prior, so that an objective function is effectively simplified; and finally, designing a quantization modulator of semantic feature-constellation point identical distribution mapping based on finite scalar quantization, and adopting a gradient straight-through estimator to solve the non-differentiable problem of the quantization process, and finally realizing end-to-end joint optimization of the system.

In accordance with an aspect of the present disclosure, there is provided a method, in an apparatus, comprising: for a channel estimate for each of at least one subset of channel estimates, obtaining a respective quantized channel estimate by performing quantization on the channel estimate using scalar quantization, the channel estimation is obtained on a channel state information reference symbol (CSI-RS) which is transmitted on a plurality of transmitting antenna ports, and the channel estimation is obtained on the channel state information reference symbol (CSI-RS) which is transmitted on the plurality of transmitting antenna ports; and transmitting channel state information (channel state information, CSI) on the basis of the quantized channel estimation.

In order to significantly reduce communication resource consumption in a wireless cooperative positioning system, the invention provides a low communication overhead position measurement and resource scheduling method, and relates to the technical field of wireless communication and positioning. The method comprises the following steps: modeling a positioning network into a directed graph; and then, dynamically evaluating the link quality by using a graph attention mechanism and calculating a communication weight between nodes, thereby realizing intelligent resource scheduling oriented to a high-quality link. Meanwhile, a shared scalar quantization codebook is designed, and a high-dimensional node state vector is compressed into a low-bit code word index for transmission, so that the communication data volume is greatly reduced. Simulation and hardware prototype experiments show that the communication overhead in the cooperative positioning process can be reduced by about 60% on the premise of keeping high positioning precision, and the problem of expansibility of high-precision positioning in a resource-limited wireless network is effectively solved.

A method and apparatus for video encoding or decoding is provided. In a method using dependent scalar quantization, the selection of a quantizer depends on the decoding of a preceding transform coefficient, and the entropy decoding of the transform coefficient depends on the selection of the quantizer. To maintain high throughput in a hardware implementation of transform coefficient entropy coding, several scalar quantizer decision schemes are proposed, in which the state transition and context model selection are based solely on the normal coding bins. For example, the state transition is based solely on the sum of the SIG, gt1, and gt2 flags, the exclusive-OR function of the SIG, gt1, and gt2 flags, or the gt1 or gt2 flags. When a block of transform coefficients is coded, the normal mode bins are coded first with one or more scan paths, and the remaining bypass coded bins are grouped with another one or more scan paths.

This document describes a Wireless Transmit / Receive Unit (WTRU). The WTRU can receive one or more reference signals and determine Channel State Information (CSI) based on the one or more reference signals. The WTRU can determine a potential vector based on compression of the CSI and perform vector quantization of the vector over multiple sets of vector quantization codebooks. The WTRU can select (1) a set of vector quantization codebooks from the multiple sets of vector quantization codebooks and (2) the number of bits used for scalar quantization based on either a CSI feedback size or a reconstruction performance criterion. The WTRU can transmit report information indicating the set of vector quantization codebooks and the quantized compressed CSI associated with the set of vector quantization codebooks.

This application concerns UE-sided beam prediction using an artificial intelligence (AI) / machine learning (ML) model. It has been suggested, in the prior art, that the UE 410 should predict the top-K best beams and report these 406 to the network together with probability information associated with each of the best beams, the latter being in a quantised form. The probability information is output from the UE-side AI / ML model. This application is concerned with the quantization of such probability information. This involves a UE transmitting, to a network device, capability information 401 relating to a quantisation scheme for beam reporting probability. The UE then “obtains” information 403, 404 relating to either a quantisation scheme or a quantization setting. This enables the UE to carry out quantisation 405 of the beam probabilities and to report the quantised probabilities 406 to a gNB 420 or other network device. The quantisation scheme is limited to a scalar quantization scheme or to use of “statistics of probability values”. The best beam may be quantised with a higher granularity than the other reported beams.

When dependent scalar quantization is used, the choice of the quantizer depends on the decoding of the preceding transform coefficient, and the entropy decoding of a transform coefficient depends on quantizer choice. To maintain high throughput in hardware implementations for transform coefficient entropy coding, several decision schemes of the scaler quantizer are proposed. In one implementation, the state transition and the context model selection are based on only regular coded bins. For example, the state transition can be based on the sum of the SIG, gt1 and gt2 flags, the exclusive-or function of the SIG, gt1 and gt2 flags, or based on only the gt1 or gt2 flag. When a block of transform coefficients is coded, the regular mode bins can be coded first in one or more scan passes, and the remaining bypass coded bins are grouped together in another one or more scan passes.

There are described techniques for encoding and decoding audio signals. A decoder, configured to generate an audio signal from a coded signal representing the audio signal, may include: a coded signal reader, configured to read the coded signal, thereby providing a plurality of indexes; a scalar dequantization module, including: a plurality of quantization index converters, each quantization index converter being configured to convert an index of the plurality of indexes onto a corresponding latent scalar value, so that a plurality of latent scalar values form a first latent audio signal representation of the audio signal; and a first learnable section to provide a second latent representation from the first latent audio signal representation; a second learnable section including at least one learnable layer and configured to generate the audio signal from the second latent audio signal representation.

The embodiment of the application provides a kind of diamond fluorescence signal low bandwidth quantization method based on error compensation, belong to quantum sensing technical field.The method comprises: obtaining the original signal to be transmitted at current time, initializing a B-bit scalar quantizer and an error memory, the original signal is summed with the quantization error saved in the error memory at previous time, to obtain correction signal;The correction signal is quantized based on the B-bit scalar quantizer, to obtain quantization output signal, and the quantization output signal is sent to receiving end;According to the correction signal and the quantization output signal, the quantization error at current time is calculated and updated to the error memory.The application compensates the quantization error at current time to the input at next time, realizes cross-time error compensation, can significantly improve the effective signal-to-noise ratio of quantum sensing signal without increasing the bit width of front-end ADC and communication bit rate.

Example embodiments of the present disclosure relate to a solution for beam reporting using quantization. In an aspect, the terminal device transmits, to a network device, capability on supported quantization scheme for reporting beam probability information associated with at least one beam predicted during inference. The terminal device receives, from the network device, at least one of quantization setting or quantization scheme configuration. The terminal device transmits, to the network device, report for at least one predicted beam and the probability information that is quantized by the terminal device based on at least one of the quantization setting or the quantization scheme configuration by using at least one of scalar quantization scheme or statistics of probability values of the at least one predicted beam. In this way, enhanced methods for configuration, determination and indication of reporting probability information with reduced overhead may be obtained.

Encoding at least one unit quaternion representing a rotation matrix used for coding a multichannel signal represented by an input point on a 4-dimensional sphere by encoding 3 spherical coordinates of the input point. The method includes sequential scalar quantization of the 3 spherical coordinates in order to obtain at most 4 candidates at the end of the sequential scalar quantization of the 3 coordinates; selecting the best candidate which minimizes a distance between the input point and the at most 4 candidates; determining the separate quantization indices resulting from the sequential scalar quantization of the spherical coordinates of the best candidate; determining a global quantization index by adding at least one item of cardinality information to the quantization index of a spherical coordinate; and coding of the global quantization index of said best candidate. A corresponding decoding method, an encoding device and a decoding device are also provided.

According to one aspect of the present disclosure, a method is provided, in an apparatus, the method comprising the steps of: obtaining a quantized channel estimate for at least a subset of channel estimates by quantizing the channel estimate using scalar quantization, wherein the channel estimate is obtained for channel state information reference symbols (CSI-RS) transmitted across a plurality of transmission antenna ports; and transmitting channel state information (CSI) based on the quantized channel estimate.

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a user equipment (UE) may receive a finite scalar quantization (FSQ) configuration of channel state information (CSI) feedback (CSF) associated with a multiple-input multiple-output (MIMO) stream. The UE may transmit, in accordance with the FSQ configuration and based at least in part on a latent feature vector, an indication of a CSF vector. Numerous other aspects are described.

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a user equipment (UE) may receive a finite scalar quantization (FSQ) configuration of channel state information (CSI) feedback (CSF) associated with a multiple-input multiple-output (MIMO) stream. The UE may transmit, in accordance with the FSQ configuration and based at least in part on a latent feature vector, an indication of a CSF vector. Numerous other aspects are described.

The application discloses a code rate delay scalable coding method and system for a vibration tactile signal, relates to the technical field of vibration tactile signal coding, and comprises the following steps: receiving a three-axis vibration tactile signal, pre-processing the three-axis vibration tactile signal to obtain a pre-processed three-axis vibration tactile signal, inputting the pre-processed three-axis vibration tactile signal into a pre-established non-stationary self-encoder model based on finite scalar quantization and attention, and outputting a normalized reconstruction frame and a difference value between the normalized reconstruction frame and a normalized frame as well as FSQ code words; calculating a residual error based on the difference value between the normalized reconstruction frame and the normalized frame, quantizing the residual error according to a quantization bit width by using a non-uniform quantization method, generating a residual error quantization index, and losslessly compressing the FSQ code words and the residual error quantization index by using entropy coding to generate an output bit stream, so that the coding performance can be improved.

The invention is suitable for the technical field of audio data processing, provides a wireless microphone audio processing method and a related device, and realizes lower delay, high stability and sound quality output acceptable to human ears for audio signals. The method is applied to a wireless audio processing system comprising a wireless microphone and wireless receiving equipment, and comprises the following steps: the wireless microphone performs feature extraction of self-adaptive dynamic framing on an original audio signal, performs weight adjustment according to a preset psychological acoustic model, and adopts finite scalar quantization mapping to obtain discrete symbols; performing context sensing entropy coding according to the characteristic sequence dependency relationship of each audio symbol in the discrete audio symbol set to obtain an audio coding stream, and transmitting the audio coding stream to wireless receiving equipment; and the wireless receiving equipment decodes and complements the audio coding stream, and synthesizes a target audio feature according to the scalar value of the complemented discrete audio symbol set and the adjusted audio weight value.

Methods, systems, and devices for wireless communications are described. In some cases, a first wireless device may use one or more reference signals to estimate a channel and generate a set of quantized bits representative of the estimated channel, where the quantization of the set of quantized bits is in accordance with a non-uniform CSI codebook that is associated with one or more spiral shapes and generated using one or more functions in accordance with the or more parameter values. The one or more parameter values may be selected by the first wireless device. Additionally, the first wireless device may transmit a report indicative of a set of quantized bits, where the set of quantized bits are quantized in accordance with the non-uniform CSI codebook, and further indicative of the one or more parameter values selected by the first wireless device.

According to an aspect of the disclosure there is provided a method, in an apparatus, the method including: for each at least a subset of channel estimates, obtain a respective quantized channel estimate by quantizing the channel estimate using scalar quantization; wherein the channel estimates are obtained on channel state information reference symbols (CSI-RS) transmitted on a plurality of transmit antenna ports; and transmitting channel state information (CSI) based on the quantized channel estimates.

The invention provides a point cloud attribute encoding and decoding method and system and a point cloud attribute compression method and system based on a lifting structure. The method comprises the steps of obtaining to-be-encoded point cloud data; obtaining a first lifting representation of the attribute signal based on the geometric coordinates using a forward first lifting structure; enabling the first lifting representation to pass through a forward second lifting structure to obtain a second lifting representation; obtaining hidden layer representation according to the second lifting structure; and quantizing the hidden layer representation by using a uniform scalar quantizer, and encoding by using the quantized hidden layer representation to obtain a point cloud attribute signal code stream. According to the method, the topological structure correlation of point cloud signals is effectively utilized, redundant information of point cloud attributes in space and channel dimensions is eliminated, and then high-performance compression of the point cloud attributes is achieved; the reversible characteristic of a lifting structure is inherited, compact multi-scale representation is generated, and powerful support is provided for high-fidelity reconstruction.

A method for encoding an input point on an n-dimensional sphere by encoding n−1 spherical coordinates of said input point. The method includes sequential scalar quantization of the n−1 spherical coordinates in order to obtain at most 2n-2 candidates at the end of the sequential scalar quantization of the n−1 coordinates, and subsequently selecting the best candidate which minimizes a distance between the input point and the at most 2n-2 candidates, and determining the separate quantization indices resulting from the sequential scalar quantization of the spherical coordinates of the best candidate and sequentially encoding the separate quantization indices of the best candidate. A corresponding decoding method, an encoding device and a decoding device are also provided.

Provided are an encoding method, an encoding device, a decoding method, and a decoding device using a scalar quantization and a vector quantization. The encoding method includes converting an input signal of a time domain into a frequency domain, generating a first residual signal from an input signal of a frequency domain by using a scale factor, performing a scalar quantization of the first residual signal, generating a second residual signal from the scalar-quantized first residual signal, performing a lossless encoding of the scalar-quantized first residual signal, performing a vector quantization of the second residual signal, and transmitting a bitstream including the lossless-encoded first residual signal and the vector-quantized second residual signal.

The application discloses a neural network joint quantization method for image classification, solves the problem that a deep neural network for image classification is difficult to be efficiently stored and run on a resource-limited embedded device, and realizes compression and acceleration of the neural network. Scalar quantization is used to convert element values of a mapping codebook obtained through product quantization into 8-bit integers. In combination with 8-bit integer activation values, fixed-point operation of the neural network is realized, and the purpose of calculation acceleration is achieved. In order to eliminate the activation value distribution error caused by different inputs, the saturation truncation method is used to reduce the quantization interval of the activation value, and 8-bit scalar quantization of the activation value is realized. Finally, a pseudo-quantity unit is inserted into the neural network, a scalar quantization error is introduced, parameters are updated through back propagation, and the error is eliminated. The application fully combines the advantages of scalar quantization and product quantization, and can significantly reduce the model size while reducing the network operation amount.

The video encoder determines scaled transform coefficients, wherein determining the scaled transform coefficients comprises scaling the transform coefficients of the block of video data according to a given quantization step. The video encoder determines scalar quantized coefficients, wherein determining the scalar quantized coefficients comprises applying scalar quantization to the scaled transform coefficients of the block. In addition, the video encoder applies a neural network that determines a corresponding set of probabilities for each corresponding transform coefficient of the block. Each probability set for each transform coefficient comprises a corresponding probability value for each possible adjustment value of a plurality of possible adjustment values. Inputs to the neural network include the scaled transform coefficients and the scalar quantized coefficients. The video encoder determines a quantization level for a particular transform coefficient of the block based on the probability set for the particular transform coefficient.

An apparatus comprising means for: receiving, from a network node, at least one ref- erence signal; determining at least one latent vector based on target Channel State Information acquired from the at least one reference signal and a history-independ- ent CSI encoder; using the at least one latent vector to provide an input into a his- tory-dependent recurrent quantizer to provide a quantized output; calculating a mean squared error between the input and the quantized output; comparing the mean squared error to a threshold level; based on the comparing, determining to op- erate in a irst mode, wherein in the irst mode, the apparatus performs: determining at least one further latent vector based on target Channel State Information acquired from the at least one reference signal and the history-independent CSI encoder; us- ing the at least one further latent vector to provide an input into a scalar quantizer to provide a full-resolution quantization output; sending the full-resolution quantiza- tion output to the network node.

In various embodiments, an encoder receives transform coefficients of prediction residues associated with video data. The encoder performs scalar quantization operations on the transform coefficients to generate first indices. The encoder performs trellis coded quantization operations on a first vector that includes the transform coefficients to generate a second vector that includes second indices. The encoder determines a first cost function value based on a third vector that includes the first indices and a second cost function value based on the second vector. The encoder sets a fourth vector that includes a plurality of quantization indices equal to either the second vector or the third vector based on whether the second cost function value is less than the first cost function value. The encoder performs entropy coding operations on the fourth vector to generate encoded video data.