185results about How to "Improve decoding accuracy" patented technology

An audio-based data indexing and retrieval system for processing audio-based data associated with a particular language, comprising: (i) memory for storing the audio-based data; (ii) a semantic unit based speech recognition system for generating a textual representation of the audio-based data, the textual representation being in the form of one or more semantic units corresponding to the audio-based data; (iii) an indexing and storage module, operatively coupled to the semantic unit based speech recognition system and the memory, for indexing the one or more semantic units and storing the one or more indexed semantic units; and (iv) a search engine, operatively coupled to the indexing and storage module and the memory, for searching the one or more indexed semantic units for a match with one or more semantic units associated with a user query, and for retrieving the stored audio based data based on the one or more indexed semantic units. The semantic unit may preferably be a syllable or morpheme. Further, the invention is particularly well suited for use with Asian and Slavic languages.

A system and method of relay code design and factor graph decoding using a forward and a backward decoding scheme. The backward decoding scheme exploits the idea of the analytical decode-and-forward coding protocol and hence has good performance when the relay node is located relatively close to the source node. The forward decoding scheme exploits the idea of the analytical estimate-and-forward protocol and hence has good performance when the relay node is located relatively far from the source node. The optimal decoding factor graph is first broken into partial factor graphs and then solved iteratively using either the forward or backward decoding schemes.

A Switching Kalman Filter Model for the real-time inference of hand kinematics from a population of motor cortical neurons. Firing rates are modeled as a Gaussian mixture where the mean of each Gaussian component is a linear function of hand kinematics. A “hidden state” models the probability of each mixture component and evolves over time in a Markov chain. Gaussian mixture models and Expectation Maximization (EM) techniques are extended for automatic spike sorting. Good initialization of EM is achieved via spectral clustering. To account for noise, the mixture model is extended to include a uniform outlier process. A greedy optimization algorithm that selects models with different numbers of neurons according to their decoding accuracy is used to automatically determine the number of neurons recorded per electrode. Closed loop neural control of external events are demonstrated using neural control of a computer curser.

The embodiment of the invention discloses an MIMO decoding method and device based on deep learning and a storage medium, and the method comprises the steps: building a training data set of MIMO decoding, and enabling the training data set to comprise a plurality of pieces of training data; Training the neural network based on the training data set to obtain a trained neural network model; And when a to-be-decoded MIMO signal is received, inputting the to-be-decoded MIMO signal into the neural network model for MIMO decoding, and then obtaining an MIMO decoding result output by the neural network model. Implementation by the invention, A neural network model used for joint MIMO detection and channel decoding is designed based on deep learning. According to the method, MIMO detection and channel decoding are taken as a joint decoding process, the approximation of an output result of a neural network model is improved through training, the overall performance of MIMO decoding is ensured,and the method has higher decoding accuracy and faster decoding speed.

The invention discloses a segmentation polarization code coding and decoding method and system based on LSC-CRC decoding. The method comprises steps of combining and separating a plurality of independent channels to obtain bit channels as many as the independent channels, obtaining the capacity of each bit channel, and obtaining the number of the fully polarized bit channels; dividing the free information bit sequence to be transmitted into the corresponding number of sub-sequences according to the number of fully polarized bit channels, subjecting each sub-sequence to polarization code decoding, and transmitting the encoded information to the corresponding bit channel; and after a receiving end receives the encoded information, performing segmentation decoding according to the LSC-CRC decoding algorithm, and finally connecting the obtained decoded sub-sequences in an end-to-end manner to obtain the decoding sequence. The channel coding and decoding method is simple to operate, the coding and decoding space complexity is reduced, and the decoding accuracy is improved.

The invention discloses an LDPC (low density parity check) decoding optimization method based on flash programming interference error perception. Along with the increase of NAND flash-memory storage density, more bit information is stored in each unit; the coupling interference along the units is high; the serious programming interference errors are caused, so that the traditional BCH code cannot ensure the data reliability; the LDPC code has error-correcting performance superior to that of the BCH code, and is applied to an NAND flash-memory storage system. The used optimized LDPC decoding algorithm has important significance. Through observation, the programming interference errors have value correlative characters; when the LDPC decoding is performed, the value correlative characters of the programming interference errors can be blended into the LDPC decoding process; the additional information is provided for bit judgment, so that the bit judgment precision is improved; the decoding convergence speed is accelerated; the LDPC decoding delay is reduced.

The invention discloses a voice self-registration method, a device, a computer device and a storage medium, which are applied in the field of artificial intelligence. The method comprises the following steps of: a voice registration request is received, wherein the voice registration request carries a registration mark; at least two guide recordings are sent to that client end according to the recording play sequence to enable the client end to play the guiding recording in sequence; registered voice data collected based on each bootstrap recording is recoreded; a target speech static decode network is used for text translation of register speech data to obtain registered text data; Text-to-speech conversion tool is used to convert the registered text data into feedback speech data, and the feedback speech data is played through the client; the confirmation voice data is acquired, and if the confirmation voice data carries the confirmation mark, the user identity registration based onthe registration text data is completed. The method can realize the process of voice self-registration and improve the flexibility and operation convenience of identity registration.

The invention discloses a Manchester decoding device capable of carrying out error correction and a method thereof. The device comprises a clock frequency divider module, a multipoint sampling module, a jump edge detection module and a decoding module. The clock frequency divider module carries out frequency division processing on a clock signal of an input apparatus and outputs the clock signal after the frequency division processing to the multipoint sampling module. The multipoint sampling module carries out multi-point sampling on a Manchester code element and outputs a sampling result to the decoding module. The jump edge detection module detects a level jump edge of each input Manchester code element and the level jump edge is taken as a reference signal of the multipoint sampling module. The decoding module carries out decoding according to a sampling result of the multipoint sampling module. A traditional Manchester decoding device based on a single point sampling method has the following technical problems that an error correction capability is poor; working is not stable; the device is easy to be interfered and reliability is not high. By using the device and the method of the invention, the above problems are well solved.

The present invention discloses a Manchester encoding signal decoding method. The method comprises the steps of receiving a Manchester encoding signal; according to the theoretical head level duration of the Manchester encoding signal, determining the sampling cycles, and setting a plurality of sampling points of same numbers in the front and back two sampling cycles of each code element evenly, and obtaining the corresponding high and low levels of each sampling point; according to the number of the high level sampling points in the front and back two sampling cycles of the code element, correcting the Manchester encoding signal; according to a Manchester encoding rule, decoding a corrected data signal. The present invention also discloses a Manchester encoding signal decoding device correspondingly. By applying the technical scheme of the present invention, the decoding accuracy can be improved.

This low density odd-even check coding control method, system and terminal device are used to solve how to raise the system throughput in the existed self-adapting control mode based on the low density odd-even check coding. This technique scheme considers fully that the decoding of LDPC is an iterative decoding, where there is a max iterative number. When the channel situation is good, only a few iterative number is enough to do decoding well. Thus some properly punching on the good channel can still keep the correct decoding with the iterative number less than the max value. Thus system throughput and frequency spectrum utilization are raised. This scheme considers that the LDPC feature of different code ratio is much different. Under the situation of punching and wrong decoding, the decoding correctness can be raised via retransmission of a few punched out data.

The invention relates to a method for correcting a position error of a permanent magnet synchronous motor rotor based on linear hall, and belongs to a method for correcting the position of the permanent magnet synchronous motor rotor. Two linear hall output signals having the difference of a 90-degrees electrical angle therebetween are acquired, fundamental components in the original signals are extracted by use of a high-precision common-frequency extraction module, and the fundamental components are fed into a rotor position angle calculator to obtain rotor angle position information. The method is characterized in that harmonics of all orders included in the original linear hall output signals are filtered out and the angle error caused by the distortions of the output signals of a linear hall position sensor can be compensated; the method is simple in algorithm, free from hardware added, easy to implement, and good in instantaneity for position signal error compensation.

Processing an analog electrical signal containing information representative of reflected light from indicia including regions of different light reflectivity, wherein the analog electrical signal contains edge transitions corresponding to boundaries between adjoining regions of different light reflectivity of the indicia. The edge transitions of at least a part of the analog electrical signal are analyzed to determine a level of blur in that part of the electrical signal. Based on the determined level of blur, one of a plurality of different techniques is selected for processing that part of the electrical signal to produce a digitized electrical signal in which transitions in the digital level of the signal correspond to boundaries between adjoining regions of different light reflectivity of the indicia.

Static transmitter stations are used to determine the position of a mobile receiver unit. Each transmitter station transmits an ultrasonic signal including a transmitter-specific phase-shifting signature. The receiver unit receives a signal and identifies a transmitter station by its signature. It uses the received signal and the identity of the source transmitter station to determine the position of the mobile receiver unit. The signature may include two patterns phase-shift-key (PSK)-encoded on respective carrier signals of the same frequency but different phase, with the patterns being offset from each other by a transmitter-specific offset. The signal from the transmitter station may also include a PSK-encoded message.

The invention provides a two-dimensional code partitioning decoding method and system. The method comprises that a received two-dimensional code image is stored in an image data backup dwell vessel; binarization is carried on the received two-dimensional code image according to gray values of pixel points; the border line of the two-dimensional code image is searched according to a position detection figure; the two-dimensional code image is further searched form the version number and a correction figure; the two-dimensional code image is partitioned into areas according to the position detection figure and the correction figure, and the bit of each area is calculated; each area is divided into bit areas according to the deflection of the position detection figure and the deflection of the correction figure; the two-dimensional code image in the image data backup dwell vessel is obtained again; the average gray value of all the pixel points in the bit areas is calculated and serves as the gray value of the bit areas; and binarization is carried out on the area according to the gray value of each bit area, and a binary two-dimensional code image is obtained; and the binary two-dimensional code image is decoded. The system is accurate in partitioning and rapid and accurate in decoding.

A composite signal including a plurality of component signal streams is processed using QR-decomposition to detect symbols and calculate a decision metric for each of the signal streams (202). Soft serial interference cancellation is performed on the composite signal based on the symbols detected for different ones of the signal streams and a ranking of the signal streams (204). The symbols are re-detected and the decision metric re-calculated for each signal stream based on soft serial interference cancellation results (206). The different decision metric calculations for at least one of the signal streams are compared to determine whether symbol detection is more accurate with or without soft serial interference cancellation (208). The signal streams are decoded using symbols detected based on soft serial interference cancellation results when symbol detection is determined to be more accurate with soft serial interference cancellation (210). Otherwise, the streams are decoded based on the symbols detected using QR-decomposition (210).

The invention relates to a novel IRIG (Inter Range Instrumentation Group)-B alternating current code intelligent decoding method. Firstly, analog low-pass filtering is carried out on an IRIG-B alternating current analog signal with the system input frequency of 1KHz to filter out the coupled environment noise of the IRIG-B alternating current analog signal in a transmission process; the IRIG-B alternating current analog signal is converted into a digital signal through an isolation sampling circuit; high-order digital filtering is carried out on the IRIG-B alternating current digital signal in an FPGA (field programmable gate array), environment noise and sampling noise are further effectively filtered, then code pattern demodulation, phase compensation, time decoding and error code modification of the IRIG-B alternating current digital signal are sequentially carried out in an intelligent decoding module, highly reliable, highly precise and intelligent decoding on the IRIG-BAC code is realized, and the method has an error code self-modification capacity, can be widely used for ground equipment and airborne equipment of a telemetering remote control system, and provides high-precision timing with strong noise resisting capacity for ground and airborne telemetering remote control systems.

The embodiment of the invention provides a decoding processing method and a decoder. The method includes the steps of receiving a log-likelihood ratio sequence sent by a decoding modulator, conducting pre-decoding on the log-likelihood ratio sequence to obtain at least two posterior probability log-likelihood ratio sequences, and conducting decoding on the posterior probability log-likelihood ratio sequences to obtain decoding results. The decoding processing method and the decoder can improve the performance gain of decoding.

The invention relates to an adaptive invisible watermark synchronous detection method based on deep learning. The method is based on an Adam optimizer, a frame comprises two U-shaped sub-networks anda CNN sub-network, and the two U-shaped sub-networks and the CNN sub-network are respectively a mark network, a detection network and a discriminator. The marking network takes square areas with fixedsizes at four corners of an extracted watermark image as input of the marking network, outputs four residual images with position information, sequentially puts the residual images back to an original image to obtain a marking image, and sends the shot marking image to the detection network during marking detection, so the detection network detects the watermark image; a mask image for displayingthe position of the residual image is outputted so as to determine four corner points of the watermark image, synchronization of the watermark image and the shot mark image is realized by using perspective transformation, and the original watermark image and the watermark mark image are distinguished by using a discriminator. According to the adaptive invisible watermark synchronous detection method based on deep learning, watermark decoding accuracy is greatly improved on the premise of ensuring the visual effect.

The invention discloses a coding chip based microarray and a preparation method thereof. In a typical embodiment, the preparation method includes: providing a carrier, which has at least one fluid accommodating cavity, in which at least one bearing surface is distributed; coating the bearing surface with an adhesive evenly; adding a selected fluid into the fluid accommodating cavity till immersionof the bearing surface with the selected fluid; depositing coding microchips dispersed in the selected fluid to the bearing surface, and performing combination with the adhesive distributed on the bearing surface; and curing the adhesive so as to fix the microchips on the bearing surface. The invention also discloses a biochemical kit based on the microarray, a detection system, a method and thelike. Compared with the prior art, the coding chip based microarray provided by the invention shows outstanding advantages in detection multiplicity, high efficiency, sensitivity and other aspects, can achieve high flux and high accuracy sample detection, also is simple in operation, and has manufacturing simplicity and economical efficiency.

The invention discloses a processing method and device of a delay between signals. The processing method comprises the following steps of: determining an error rate of each of multiple paths of serial digital signals in N sampling clocks, wherein each sampling clock in the N sampling clocks is the sum of a recovery clock and N interpolating phases, and the N interpolating phases are preset in a clock unit; determining the interpolating phase corresponding to each path of serial digital signal according to the error rate, wherein the position of each sampling clock is preset in a clock unit; and adjusting a clock of each path of serial digital signal by using the interpolating phase corresponding to each path of serial digital signal. According to the processing method and device disclosed by the invention, the data transmission reliability is improved.

The invention discloses a DNN decoding method of an SCMA system and decoding communication equipment. The DNN decoding method of the SCMA system comprises the steps of S1, building the SCMA system, and obtaining a training sample data set; S2, establishing an SCMA decoder model based on a deep neural network; S3, training the SCMA decoder model; And S4, deploying an SCMA decoder model, and decoding the SCMA signal through the SCMA decoder model. According to the method, the SCMA decoding accuracy can be improved on the premise that the complexity of the SCMA decoder is not increased, and compared with a traditional SCMA decoder based on an MPA algorithm, the SCMA decoder based on the DNN has the advantages that the calculation complexity and the decoding bit error rate are improved.

The invention discloses a multi-scale information enhanced binocular convolutional neural network saliency image detection method, and relates to the technical field of neural networks. The method comprises the steps of constructing an end-to-end convolutional neural network binary classification training model, inputting original scene image data in a training set into the convolutional neural network binary classification training model for training, and obtaining a final saliency prediction image; and repeatedly calculating a loss function value between the final saliency prediction image and the real saliency detection image set to obtain a trained convolutional neural network binary classification training model, and inputting a scene image to be subjected to saliency detection into the convolutional neural network binary classification training model to obtain a predicted saliency detection image. The saliency detection image generated by using the method is clear in boundary andcomplete in structure, and the problems of structural loss and incomplete details of the saliency detection image generated by using the existing method are solved.

The present invention provides a stream decoding method and device.The method includes following steps:receiving the stream, the stream including the first protocol form stream or the second protocol form stream; decoding according to the decoding form of the second protocol form if judging the stream being the first protocol form stream by the stream mark and detecting the stream mark under the decoding course of the second protocol form; judging the amount of the stream marks of the first protocol form whether exceeding the preset threshold or not if judging the stream being the second protocol form stream by the stream mark and detecting the stream mark of the first protocol form in the course of decoding form of the second protocol form; if exceeding, decoding according to the first protocol form. The embodiment of the invention improves the accuracy rate of the decoding stream and ensures the decoding priority of the decoding stream.

The invention provides a block decoding method and system for a two-dimensional code. The block decoding method includes: storing a received two-dimensional code image in an image data backup register; binarizing the received two-dimensional code image; searching for a boundary line of the two-dimensional code image; segmenting each of the regions into bit regions each of which corresponds to one bit data; re-acquiring the received two-dimensional code image stored in the image data backup register; calculating an average gray-scale value of all the pixels in the bit region as a gray-scale value of the bit region; calculating a gray-scale threshold of each of the bit regions n using the Otsu method; binarizing the bit region so as to obtain a binarized two-dimensional code image; and decoding the binarized two-dimensional code image. In the system, the two-dimensional code image is segmented precisely, thereby effecting quick and accurate decoding.

Disclosed in the invention is an automatic repeat request (ARQ) controlling method of multi-media data. The ARQ controlling method comprises the following steps: obtaining a lost multi-media data packet, priority information of the lost multi-media data packet, and a current network condition; and employing a corresponding re-transmission strategy to re-transmit the lost multi-media data packet. In addition, the invention also discloses an ARQ controlling system for the multi-media data. A problem that correct decoding relying on a multi-media data packet with the high priority level can not be carried out due to abandoning of the multi-media data packet with the high priority level when ARQ controlling is carried out can be solved.

Static transmitter stations are used to determine the position of a mobile receiver unit. Each transmitter station transmits an ultrasonic signal including a transmitter-specific phase-shifting signature. The receiver unit receives a signal and identifies a transmitter station by its signature. It uses the received signal and the identity of the source transmitter station to determine the position of the mobile receiver unit. The signature may include two patterns phase-shift-key (PSK)-encoded on respective carrier signals of the same frequency but different phase, with the patterns being offset from each other by a transmitter-specific offset. The signal from the transmitter station may also include a PSK-encoded message.

The invention discloses a method and a system for demodulating a frequency shift keying modulation signal, and relates to the field of frequency shift keying modulation and demodulation. The method comprises the following steps of processing data sent by an external radio frequency system to obtain two paths of intermediate-frequency I / Q data, and removing direct current offsets; carrying out zerointermediate frequency processing through a numerically controlled oscillator; limiting out-of-band noise of zero-intermediate-frequency I / Q data through low-pass filtering, and becoming phase data after time phase conversion is carried out; limiting out-of-band noise of the phase data through low-pass filtering, and generating a bit code stream after differential subtraction and decision operation; comparing the generated bit code stream with an address code, and when the number of comparison results is smaller than a preset maximum number, outputting frame synchronization pulse; after the bit code stream is subjected to frequency offset calculation, outputting a frequency offset code; and after the output frequency offset code is demodulated, outputting a final demodulation result bit code stream and carrying out clock synchronization processing simultaneously. The method and the system are suitable for frequency shift keying modulation and demodulation.

The invention discloses a joint optimization method for dimensionality reduction and quantification of a communication signal for object state estimation. The joint optimization method specifically comprises the following steps: (1) reducing the dimensionality of a system state through a state linear dimensionality reduction method to realize dimensionality limit for independent parallel signal channel transmission data; (2) predicating through a multi-level quantization innovation kalman filter to obtain a predicted value of a signal and furthermore obtain the innovation of the signal; (3) quantifying the innovation under the restraint of transmission power and bandwidth by taking least mean-square estimation as the standard by using a joint-design state linear dimensionality reduction method and a time-invariant symmetry quantization method; and (4) filtering the received innovation through the multi-level quantization innovation kalman filter at a signal receiving end to realize object state estimation. By using the joint optimization method, a transmission signal meets communication restraint, and meanwhile, a state estimator at the signal receiving end has higher estimation precision, therefore, the joint optimization method is an effective signal processing method.

The invention discloses a method for embedding a blockchain ledger in audio digital watermarking. The method comprises the following steps: I, selecting separators, block sizes, head structures of blocks and transaction details as ledger information and watermarking information and combining the information so as to achieve quantization and coding of watermarking containing a blockchain data; II,reducing dimensions of the blockchain ledger information and watermarking information obtained from quantization and coding, transforming a frequency domain of original audio and implementing quantization and coding once again, so that new matrix coefficient coding is formed, and implementing frequency domain inverse transformation so as to obtain final audio watermarking containing the blockchaininformation, so that blockchain watermarking embedding is completed; and III, updating watermarking information in the blockchain by updating a data value in the blockchain information, so as to facilitate tracking and to acquire a performance that can not be arbitrarily modified. According to the method provided by the invention, quantized information of the blockchain ledger, as one part of thewatermarking information, is embedded into a transformation domain of the audio signal, so that the integrity of the watermarking can be kept in a transfer process and a bit error rate in data extracting can be reduced.

The invention discloses an improved segmented stepped phase encoding three-dimensional measurement method. The method is composed of four key parts including an intermittent stepped phase encoding principle, a segmented phase encoding principle, a complementary fringe order correction principle and a three-dimensional measurement principle. The method has the advantages that the phase instead of the intensity coding mode is used, and the method is insensitive to factors such as contrast of the surface of a measured object, ambient light and camera noise and is good in robustness so that the method can measure objects with different surface brightness degrees; the number of code words is increased through the idea of segmentation, and the height of adjacent steps is increased by designing intermittent step phases so that the measurement precision can be improved, the stripe level after two times of left-right phase shift is used for complementing hopping of the original stripe level, the decoding accuracy is greatly improved, and the method has a certain application prospect in measuring the three-dimensional morphology of a complex industrial product.