364 results about "Viterbi algorithm" patented technology

Techniques are provided for reducing feedback while maintaining performance in a MIMO-OFDM system. The disclosed techniques employ finite-rate feedback methods that uses vector quantization compression. The disclosed methods / techniques generally involve: receiving a plurality of symbols from a plurality of sub-carriers at a receiver; selecting a plurality of indices of codewords corresponding to a codebook of pre-coding weighting matrices for the sub-carriers based on vector quantization compression of the codewords; and transmitting the selected indices over a wireless channel to a transmitter. Finite state vector quantization feedback makes use of a finite state vector quantizer (FSVQ), which is a recursive vector quantizer (VQ) with a finite number of states. In finite state vector quantization feedback, optimal precoding matrices (beamforming vectors) are selected sequentially across subcarriers. In a trellis-based feedback method, the optimal precoding matrices are selected at the same time for all subcarriers by searching for the optimum choice of matrices along a trellis using the Viterbi algorithm (dynamic programming).

A Viterbi decoding system interprets bits in received QAM constellations as many-valued parameters rather than binary valued parameters. It performs the Viterbi algorithm using these many-valued parameters to provide results superior to hard decision decoding. Rather than applying a hard 0-1 function to the QAM data, the system uses a non-stepped linear or curved transfer function to assign values to the bits. In another aspect, a system differentiates between data bits based on their estimated reliability, giving more emphasis to decoding reliable bits than unreliable bits using any of a variety of techniques. By differentiating between god and bad bits and de-emphasizing or ignoring unreliable bits, the system can provide a significant reduction in uncorrectable errors and packet loss.

A repeatable run-out (RRO) detector employs one or more digital interpolators to interpolate asynchronous sample values representing an RRO address mark (AM) and RRO data, an asynchronous maximum-likelihood (AML) detector to detect the RRO AM, and a RRO data decoder to decode the RRO data. The AML detector employs an AML algorithm, such as a Viterbi algorithm, to detect the series of peaks of the RRO AM based on detection of the entire sequence of observed peaks. AML detection selects one of either the asynchronous or interpolated sample sequences that are closest in distance to the ideal RRO AM sample sequence. Once the RRO AM is detected, the AML detector provides a RRO AM found signal as well as the selected one of the sample sequences having the best phase for detecting and decoding the RRO data.

Joint equalization and decoding techniques to eliminate the division operations in the Frequency domain Equalizer (FEQ) of an Orthogonal Frequency Division Modulation (OFDM) receiver by incorporating the magnitude of the channel impulse response estimates into the decision metrics in the decoder. This includes methods for both hard-decision symbol-by-symbol detection and soft-decision decoding using the Viterbi algorithm. Further, Channel State Information (CSI) is incorporated to improve the performance of the receiver. The disclosure also introduces a computationally efficient bit-by-bit piecewise approximation technique incorporating CSI to implement decoder decision metrics. Finally, an efficient implementation technique for multiple receiver antennae using Maximum Ratio Combining (MRC) and Viterbi decoding with CSI is disclosed.

The invention discloses a method for identifying a walking mode and tracing a track of a pedestrian in a room. The method comprises the steps: firstly, defining a walking mode of the pedestrian in the room by using directions and speeds as the classifying basis according to the characteristic of a terrain under an indoor environment; then establishing a mathematical modeling for the indoor walking mode by applying a hidden Markov model (HMM) through using acquired acceleration sensor and geomagnetic sensor data; because of the interference of a metal object and the influence of body joggling in the walking process, a track primarily obtained by an indoor positioning system through dead reckoning possibly has large errors, and in the method, calculating the probability of each walking mode by using a viterbi algorithm of the HMM, and finding one group of walking modes with highest possibility; finally, matching an original location to a corresponding walking mode according to an identification result of the walking mode so as to correct an original track, and obtain a walking correcting track of the pedestrian in the room. According to the method, the precision of indoor positioning and track tracing can be remarkably improved.

Various systems and methods for error reduction in a digital information system are disclosed herein. As one example, a digital storage system is provided that includes a soft output Viterbi algorithm channel detector operable to receive an encoded data set, and to provide a hard and a soft output representing the encoded data set. The hard and the soft output from the soft output Viterbi algorithm channel detector are provided to a single parity row decoder that provides another hard output that is an error reduced representation of the encoded data set. The encoded data set is additionally provided from the buffer to another channel detector via a delay element. The hard output from the single parity row decoder and the time shifted encoded data set are provided to coincident with each other to another channel detector.

In an error-correcting decoder, in which an input digital signal including reliability information is decoded by using a Viterbi algorithm as a first decoding process and a final decoded result is obtained by block-code decoding as a second decoding process, a flag signal is added to a location where a value of reliability of path metric determined by the Viterbi algorithm is lower than a threshold, as an original flagged location. A flag signal adding unit continuously adds flag signals to locations, from the original flagged location to locations preceding the originally flagged location, after back tracing. The flagged locations are then regarded as erasure locations in the block-code decoding process.

The invention provides a block-encoded modulation scheme using multi-level partitioning techniques. This scheme is made transparent to phase ambiguities of ±pi / 2 and of pi, by means of differential encoding and appropriate mapping, it is applicable to 16-QAM modulation, and it has theoretical encoding gain that is optimal for the rate of the code. The decoder associated with this scheme uses the Wagner algorithm which is much less complicated to implement than the Viterbi algorithm or than the Reed-Solomon algorithm.

The invention relates to a vehicle GPS data map matching method based on a hidden markov model. The method comprises the following steps: acquiring route data from a shpefile electronic map; extracting original vehicle trajectory data, and preprocessing vehicle GPS data; taking a road section with a certain distance of each GPS observation point as a candidate road section; calculating the observation probability of each GPS point and transition probability of an adjacent candidate road section on the basis of the hidden markov model; and calculating an optimal matching trajectory by using a viterbi algorithm. The vehicle GPS data map matching method is based on the hidden markov model, by consideration of the positions of the GPS points, speed and direction, topology of a road network and the associated information between trajectory points and the road network, new observation probability and new transition probability are raised, and therefore, map matching accuracy is improved further.

The invention discloses an industry spelling mistake checking method based on user feedback. According to the industry spelling mistake checking method based on user feedback, spelling mistake checking is carried out on English text by using an N-gram method and a user dictionary which is designed in a classified mode, recommendation of correct words is accomplished by searching for a large corpus database, and thus checking of spelling mistakes related to a user is achieved. The N-gram method serves as a basic method for natural language processing, and the mistakes in the text are checked according to the characteristics of words or statements and statistical information in a corpus; recommended words which are most related to wrong words in the text input by the user are selected through cooperation between the user dictionary designed in the classified mode and statistical data of the corpus according to historical information of the user at present; the database is searched for a word chain with the largest conditional probability product by using the Viterbi algorithm, and computational efficiency of a hidden Markov model in the large corpus and use efficiency of the statistical information in the database are improved.

The invention provides a method for identifying geographical name entities. The method comprises the steps that address texts are input and pre-processing is conducted on the address texts; address segmentation is conducted on the address texts according to dictionary metadata; address labeling is conducted on segmentation results, and an optimal address level labeling sequence is obtained; according to the context, the labeling sequence is revised, and an optimal labeling result is output. According to the dictionary metadata, the address segmentation is conducted on the address texts, then corresponding geographical name categories are labeled on the segmentation result according to a geographical name category definition table, the labeling sequence is optimized through a Viterbi algorithm, the labeling sequence is revised according to the context, and the final labeling result is obtained. Therefore, the geographical name entity identification result is accurate, and practicability is high. In addition, the invention provides a system for identifying the geographical name entities.

The invention provides a character recognition method based on deep learning. The character recognition method based on deep learning comprises the steps that a deeper multilayer convolution neural network structure is designed, and each character acts as a class; the convolution neural network is trained by using a back propagation algorithm so as to recognize the single character, the target function of the network is minimized in a supervised way and a character recognition model is obtained; and finally the most possible words are found out of the dictionary by adopting a viterbi algorithm according to the existing recognized characters. One input is given, candidate characters are obtained by performing sliding window scanning firstly and then the most possible words are found out of the candidate characters in testing. The character features are learned by utilizing the deeper convolution neural network, the method has robustness in character color, size, illumination and fuzziness, and relatively high accuracy of character recognition and word recognition can be maintained.

The invention relates to a non-stationery vibration signal instantaneous frequency estimation algorithm in a start and stop period of rotating machinery; firstly, an experiment table working model of a rotor of the rotating machinery is built, a non-stationery vibration signal sensor for measuring vibration signal, and a photo-electric sensor for measuring the rotating speed of a reference axis are arranged along the horizontal and vertical direction of the reference axis; after the obtained signal is analyzed by an order analysis software of an upper computer through a dynamic signal analyzer, STFT time-frequency spectrum containing multi-level component is obtained; the working frequency of the rotating machinery is used as an estimated starting frequency point, according to the requirements of sampling frequency and calculation accuracy, the frequency obtained by the STFT time-frequency analysis is equally divided into M groups, the start and stop period is equally divided into N time points for building grid meshes of N time points and M groups of frequency points; the route of minimum-deviation frequency point from the start point to the stop point is computed through Viterbi algorithm, after fitting is carried out, the instantaneous frequency estimation function value of the reference axis of the non-stationery vibration signal of the rotating machinery in the start and stop period is obtained.

A receiver generates log-likelihood-ratio-based soft bit metrics of precoded quaternary continuous phase modulation signals using four state-constrained trellises and a streamlined maximum likelihood sequence estimation Viterbi algorithm requiring no survivor state storage elements for a preferred error correction-coded quaternary Gaussian minimum shift keying communication system employing reduced-complexity pulse-amplitude modulation matched-filtering and soft-decision decoding.

A fast variational on-line learning technique for training a transformed hidden Markov model. A simplified general model and an associated estimation algorithm is provided for modeling visual data such as a video sequence. Specifically, once the model has been initialized, an expectation-maximization (“EM”) algorithm is used to learn the one or more object class models, so that the video sequence has high marginal probability under the model. In the expectation step (the “E-Step”), the model parameters are assumed to be correct, and for an input image, probabilistic inference is used to fill in the values of the unobserved or hidden variables, e.g., the object class and appearance. In one embodiment of the invention, a Viterbi algorithm and a latent image is employed for this purpose. In the maximization step (the “M-Step”), the model parameters are adjusted using the values of the unobserved variables calculated in the previous E-step. Instead of using batch processing typically used in EM processing, the system and method according to the invention employs an on-line algorithm that passes through the data only once and which introduces new classes as the new data is observed is proposed. By parameter estimation and inference in the model, visual data is segmented into components which facilitates sophisticated applications in video or image editing, such as, for example, object removal or insertion, tracking and visual surveillance, video browsing, photo organization, video compositing, and meta data creation.

The invention relates to a BBS user abnormal behavior auditing method based on Hidden Markov theory, belonging to the technical field of computer and information safety. The method comprises the following steps: reading auditing data from a database to obtain behavior observed sequence to be detected; reading trained model parameters from a Hidden Markov model database; calculating a state value sequence by utilizing a Viterbi algorithm; calculating the observed value sequence probability by forward and afterward algorithms, comparing the observed value sequence probability with the preset threshold, and giving an alarm if the observed value sequence probability is lower than the threshold, otherwise, belonging to normal behavior, correcting the Hidden Markov model parameter and storing the Hidden Markov model parameter into a parameter file of the Hidden Marko model. The BBS added with audit tracking helps system managers to prevent systems and resources from being damaged by illegal authorized users and provides help when data is recovered. The audit tacking can realize associated targets of safety, comprising personal function, event reconstruction, intrusion detection and fault analysis, and greatly improves the safety of the system.

Aspects of the present invention include systems and methods for segmentation and recognition of action primitives. In embodiments, a framework, referred to as the Continuous Linear Dynamic System (CLDS), comprises two sets of Linear Dynamic System (LDS) models, one to model the dynamics of individual primitive actions and the other to model the transitions between actions. In embodiments, the inference process estimates the best decomposition of the whole sequence into continuous alternating between the two set of models, using an approximate Viterbi algorithm. In this way, both action type and action boundary may be accurately recognized.

Being adopted for modulation method in the invention, waveform storing quadrature modulation includes steps: decomposing Phase Phi(t, Alpha) to Thetan and Theta (t, Alpha); building a corresponding look up table, storing their sine value and cosine value; through look up table to obtain sine value and cosine value of Thetan and Theta (t, Alpha); then through difference operation to obtain baseband signal. Being adopted for demodulation method in the invention, simplified noncoherent viterbi algorithm simplifies state of CPM signal as Sigman'; receiver processes the intermediate frequency signal to obtain complex baseband signal GammaB (t); the baseband signal multiplied by a delayed and shifted signal obtains baseband differential signal; according to grid chart, calculating branch measure of each state Sigma n+1'; then estimating not simplified ML state Sigma' n+1 (Sigma' n+1), and tracing back to output code element. Features are: easy of calculation and implementation and robustness.

The invention relates to an urban air quality grade predicting method based on multi-field characteristics. The method comprises a data preprocessing stage, a training stage and a predicting stage. In the data preprocessing stage, mesh generation is carried out to obtain a training sample, and a training data set is obtained. In the training stage, the optimal parameter of a conditional random field model is estimated, an optimal model shown in the specification is obtained, and the conditional random field model is output. In the predicting stage, the maximum output sequence, shown in the specification, of the conditional probability P(Y / X<f>) is obtained by combining the viterbi algorithm and the conditional probability model shown in the specification and obtained in learning of the training stage, and a prediction result is output. The AQI grade can be predicted more accurately, the AQI grade prediction in the following time quantum of all stations is carried out, and the requirement for selecting the time quantum with the high air quality to go out and take activities of people is met.

The invention relates to a method for improving rejection capability of a speech recognition system. The method comprises the following steps of collecting various types of noise data; classifying according to the noise types; for different types of noise, respectively training GMMs (Gauss mixed model); assembling various types of GMMs into an integral absorption model; training a statistic language model by various types of relatively random texts, and then establishing a recognition network by WFST (weighted finite state transducer) technique, which is called as an absorption network; connecting the absorption network, the absorption model and an original decoding network in parallel to form a new decoding network; enabling the input original audio frequency to pass endpoint detection and a feature extraction module, so as to generate feature vectors; and competing the feature vectors in the three parts of the decoding network according to an Viterbi algorithm, so as to generate a final recognition result, and effectively reject the noise and an out-of-vocabulary condition. The method has the advantage that on the premise of balancing the recognition efficiency, the effect of rejecting the out-of-vocabulary condition and the invalid input is well realized.

Methods and apparatus are disclosed, such as those involving a flash memory device that includes an array of memory cells. One such method includes detecting values of charges stored in selected memory cells in the memory cell array. The method also includes processing the detected values in accordance with a Viterbi algorithm so as to determine data stored in the selected memory cells. In one embodiment, the flash memory cell array includes word lines and bit lines. Detecting the values of charges includes detecting values of charges stored in a selected row of memory cells by selecting one of the word lines. The Viterbi algorithm provides correct data where inter-signal interference between the cells affects the accuracy of read data. For example, the Viterbi algorithm can be used to supplement error correction codes (ECC).

A recognition ambiguity resolution method of a Chinese named entity belongs to the field of named entity extraction; the method comprises the following steps: (1) the invention adopts a feature induction to reduce a feature number, i.e., a classifier provided by the invention automatically selects a meaningful feature through training and learning; (2) after selecting the feature, an N-BEST best marking sequence is selected from the CRF (Conditional Random Field) model learning through a Viterbi algorithm, i.e., N marking sequences with maximum probability of an observation sequence are selected; (3) in consideration of an occurrence frequency and a word length of the Chinese named entity, the method disclosed by the invention adopts a modified greedy algorithm to implement the ambiguity resolution, thus the entity marking sequence is obtained.

Aspects include methods, computer readable storing instructions for such methods, and systems for processing text strings such as URLs that comprise patterns of parameters and values for such parameters, delimited in a site-specific manner. Such aspects provide for accepting a number of text strings that are expected to have a common delimiting strategy, then deeply tokenizing those text strings to arrive at a set of tokens from which are selected anchor tokens used to form patterns having the anchor tokens separated by wildcard portions for recursive processing. The patterns formed can be mapped to a tree of nodes. Information concerning relationships between nodes and between tokens within a given node, as well as other heuristics concerning which tokens are parameters and which are values can be used as observed events for producing probabilities that certain tokens are parameters or values, using a dynamic programming algorithm, such as a Viterbi algorithm.

Joint equalization and decoding techniques to eliminate the division operations in the Frequency domain Equalizer (FEQ) of an Orthogonal Frequency Division Modulation (OFDM) receiver by incorporating the magnitude of the channel impulse response estimates into the decision metrics in the decoder. This includes methods for both hard-decision symbol-by-symbol detection and soft-decision decoding using the Viterbi algorithm. Further, Channel State Information (CSI) is incorporated to improve the performance of the receiver. The disclosure also introduces a computationally efficient bit-by-bit piecewise approximation technique incorporating CSI to implement decoder decision metrics. Finally, an efficient implementation technique for multiple receiver antennae using Maximum Ratio Combining (MRC) and Viterbi decoding with CSI is disclosed.

The invention discloses a BILSTM-CRF product name identification method based on self-attention. The identification method is characterized by comprising three parts of semi-automatic product title data labeling, model construction and training and model use. The semi-automatically labeling product title data part establishes an iterative process of preliminary labeling, learning, label prediction, manual correction, learning and label prediction, the model construction and training part includes performing N-dimensional dense vector coding on each word, inputting the N-dimensional dense vector coding into a BiLSTM layer to obtain text sequence characteristics, and obtaining a label probability of each word by utilizing a Softmax classification layer; extracting text local features by using the CRF layer, and training a model; the model use part is used for extracting text features and obtaining probability of all labels by utilizing a classification layer; and obtaining a corresponding tag by using a Viterbi algorithm, thereby identifying the product name. According to the method, the labor cost is greatly reduced, and the model accuracy and robustness are improved.

The invention relates to an intelligent word segmentation method based on a hidden Markov model. The method comprises the following steps of (1) building a parameter Lambda<0>=(N, M, L, Pi, A, B<1>, B<2>) of the hidden Markov model; (2) determining a state set Theta in an article; (3) abbreviating Lambda<0>=(N, M, L, Pi, A, B<1>, B<2>) as Lambda=(Pi, A, B<1>, B<2>) after determining N, M and L; (4) carrying out word segmentation on a large amount of articles by a mechanical word segmentation method through applying computer languages, and then marking the states of the articles by a computer to further form an initial Pi matrix, an A matrix, a B<1> matrix and a B<2> matrix; (5) carrying out article training on the formed initial A matrix, the B<1> matrix and the B<2> matrix by using a BW algorithm, and revaluating according to a BW algorithm revaluation formula to obtain a new Pi matrix, a new A matrix, a new B<1> matrix and a new B<2> matrix; and (6) carrying out Chinese word segmentation by using a viterbi algorithm according to a new parameter of the hidden Markov model (please see the abstract), dividing the article into a plurality of sentences according to punctuation symbols, and carrying out Chinese word segmentation on each sentence, thereby obtaining the article after word segmentation. By the intelligent word segmentation method, accurate and high-efficiency word segmentation can be carried out on a large amount of Chinese texts.

Disclosed is a technique for mitigating the effect of an in-band interferer in Orthogonal Frequency Division Multiplexing (OFDM) wireless or wired networks that employ soft decision Viterbi decoder in the physical layer. The technique uses an independent estimation of the signal quality, which is passed to the Viterbi algorithm decoder so that it is able to discard the corrupted subcarriers. This improves the error correction capability of the Viterbi algorithm decoder in a receiver, which leads to fewer retransmissions and into higher information throughput.

The invention discloses a rotary machine instantaneous rotation speed estimation method based on vibration signal synchronous compression transformation. The rotary machine instantaneous rotation speed estimation method comprises the following steps of 1 obtaining a vibration signal in the rotary machine operation process; 2 conducting frequency shift treatment on the measured vibration signal, 3 conducting synchronous compression continuous wavelet transformation on the vibration signal subjected to the frequency shift treatment to obtain the time-frequency distribution of the vibration signal subjected to the frequency shift treatment, 4 utilizing a Viterbi algorithm to extract first-order instantaneous frequency components of the vibration signal subjected to the frequency shift treatment from the obtain time-frequency distribution, and 5 utilizing the extracted first-order instantaneous frequency to recover calculation so as to obtain the rotary machine instantaneous rotation speed. The rotary machine instantaneous rotation speed estimation method adopts a frequency shift algorithm and the synchronous compression continuous wavelet transformation to process the signal, achieves accurate estimation of the instantaneous frequency of the vibration signal, utilizes the Viterbi algorithm to achieve accurate extraction of the instantaneous frequency and can accurately extract the instantaneous rotation speed of a rotary machine which cannot directly measure the instantaneous rotation speed through the vibration signal.

The invention provides a method and apparatus for estimating flat fading channel in CDMA communication system, said method is implemented by using an adaptive forward prediction technique based on lattice filter and maximum likelihood technique of Viterbi algorithm. The adaptive lattice filter is used to carry out prediction of LS criteria on channel fading, and a maximum likelihood detection technique completes Viterbi algorithm in accordance with a channel fading value obtained by the prediction, thus obtaining final estimation and decision about the transmitting signals. The present invention has the advantages that it can obtain accurate result for channel estimation and sequence decision when it operates in the fast fading channel, and overcome fast fading influence due to motion speed up of mobile station, thereby satisfying mobile station speed and corresponding receiving performance required in 3G mobile communication.