529 results about "Logit" patented technology

Locating an eye includes generating an intensity response map by applying a 3-rectangle filter and applying K-mean clustering to the map to determine the eye. Locating an eye corner includes applying logarithm transform and grayscale stretching to generate a grayscale eye patch, generating a binary map of the patch by using a threshold based on a histogram of the patch, and estimating the eye corner by averaging coordinates weighted by minimal eigenvalues of spatial gradient matrices in a search region based on the binary map. Locating a mouth corner includes generating another intensity response map and generating another binary map using another threshold based on another histogram of the intensity response map. Locating a chin or a cheek includes applying angle constrained gradient analysis to reject locations that cannot be the chin or cheek. Locating a cheek further includes removing falsely detected cheeks by parabola fitting curves through the cheeks.

Fingerprints are bit strings extracted from a media signal (e.g. an audio or video clip) to identify said media signal. Typically, they are derived from a perceptual property of the signal, for example, the spectral energy distribution of an audio fragment or the luminance distribution of a video image. A method and arrangement for extracting a fingerprint is here disclosed which is robust with respect to shifts of the perceptual property. Such shifts occur, inter alia, when the fingerprint is derived from a logarithmically mapped spectral energy distribution of an audio signal and said audio signal is subjected to speed changes. According to the invention, the fingerprint is not derived from the perceptual property as such, but from its auto-correlation function.

In one embodiment of the present invention, a decoder receives a dependency tree as a source language input and accesses a set of statistical models that produce outputs combined in a log linear framework. The decoder also accesses a table of treelet translation pairs and returns a target dependency tree based on the source dependency tree, based on access to the table of treelet translation pairs, and based on the application of the statistical models.

Methods and apparatus are provided for measuring the performance of a read channel. A number of detection techniques, such as SOVA and maximum-a-posteriori (MAP) detectors, produce a bit decision and a corresponding reliability value associated with the bit decision. The reliability value associated with the bit decision may be expressed, for example, in the form of log likelihood ratios (LLRs). The reliability value can be monitored and used as a performance measure. The present invention provides a channel performance measure that generally correlates directly to the BER but can be collected in less time.

An integrated maximum a posteriori equalization and turbo product coding (IMAP-TPC) system for optical fiber communications systems (OFCS) is provided that uses probabilistic characterization of the electrical current in the presence of inter-symbol interference (ISI) and noise to compensate their effects and improve the bit error rate. In the IMAP-TPC system, turbo product code (TPC) decoding is integrated with a symbol-by-symbol maximum a posteriori (MAP) detector. The MAP detector calculates the log-likelihood ratio of a received symbol using the conditional electrical probability density information, and hence obtains a much more accurate reliability measure than the traditional measure used in the TPC decoder.

A novel and useful apparatus for and method of multiple input multiple output (MIMO) detection for use in MIMO based communication systems. The mechanism of the invention performs a simplified tree search utilizing a single stage expansion of the most likely first symbol candidates, in the case of a 2×2 MIMO system. The invention also provides a refinement mechanism that is operative to significantly improve the log likelihood (LLR) of the list of candidates. To improve the LLR, the mechanism applies refinements rounds to generate additional candidates for both first and second detected symbols.

The present invention includes a method for determining interval values of seismic quality factor, Q, from seismic data. Seismic data is recorded and preprocessed as necessary. Estimates of amplitude spectra are determined from the seismic data. Logarithms are taken of the amplitude spectra and weights derived from the amplitude spectra. Interval values of seismic quality factor, Q, are determined by performing a weighted fit to the log-amplitude spectra with a function that is parameterized by an initial wavelet, an attenuation profile (1/Q) and an absolute-scaling profile.

An apparatus and method are provided for normalizing input soft metric to a channel decoder in a wireless communication system. A demapper generates soft metric using an in-phase component (X_k) and a quadrature component (Y_k) of a received modulated symbol (R_k), a channel fading coefficient (g_k) and a constant value (c) defined by a modulation order of the received modulated symbol. A normalizer receives the soft metric, computes a normalized log likelihood ratio (LLR) by multiplying the soft metric by a ratio of the constant value to a noise variance value, transforms the normalized LLR into a desired range and a desired number of bits, and outputs an input LLR of the channel decoder.

A forward error correction method for decoding coded bits generated by low density parity check matrixes. The method comprises converting each of the coded bits into a log likelihood ratio (LLR) value, and applying the converted values to variable nodes; delivering messages applied to the variable nodes to check nodes; checking a message having a minimum value among the messages, and determining a sign of the message having the minimum value; receiving messages updated in the check nodes, adding up signs of the received messages and a sign of an initial message, applying a weighting factor of 1 when all signs are identical, and when all signs are not identical, updating a message of a variable node by applying a weighting factor; determining LLR of an initial input value; and hard-deciding values of the variable nodes, performing parity check on the hard decision values, and stopping the decoding when no error occurs.

A method for detecting a timeout condition for a data item (e.g., a request) within the process (e.g., within an arbitration process) includes maintaining a current time as a first N-bit binary number (A). An event time of an occurrence of an event pertaining to the data item within the process is recorded and stored as a second N-bit binary number (B). A predetermined time limit, expressed as a non-negative integer K, is configured. K is less than N and K is a logarithm base 2 of the predetermined time limit. A timeout condition pertaining to the data item is detected when a difference between the current time and the event time exceeds the predetermined time limit. The detection of the timeout condition is performed utilizing a single-stage operation. This single stage operation may include computing A (current time)−B (event time) modulo 2ⁿ≧2^k.

The present invention provides systems and methods for signal detection and enhancement. The systems and methods utilize one or more discriminative classifiers (e.g., a logistic regression model and a convolutional neural network) to estimate a posterior probability that indicates whether a desired signal is present in a received signal. The discriminative estimators generate the estimated probability based on one or more signal-to-noise ratio (SNRs) (e.g., a normalized logarithmic posterior SNR (nlpSNR) and a mel-transformed nlpSNR (mel-nlpSNR)) and an estimated noise model. Depending on the resolution desired, the estimated SNR can be generated at a frame level or at an atom level, wherein the atom level estimates are utilized to generate the frame level estimate. The novel systems and methods can be utilized to facilitate speech detection, speech recognition, speech coding, noise adaptation, speech enhancement, microphone arrays and echo-cancellation.

A single-ended blur detection probe and method with a local sharpness map for analyzing a video image sequence uses two sets of edge filters, one for “fast edges” and the other for “slow edges.” Each set of edge filters includes a horizontal bandpass filter, a vertical bandpass filter and a pair of orthogonal diagonal filters where the frequency response of the fast edge filters overlap the frequency response of the slow edge filters. The video image sequence is input to each filter of each set, and the output absolute values are combined with weighting factors to produce a slow edge weighted sum array and a fast edge weighted sum arra. The respective weighted sum arrays are then decimated to produce a slow edge decimated array and a fast edge decimated array. The ratio of the maximum difference value between the decimated arrays and the maximum value from the fast edge decimated array, weighted by an appropriate factor, produces a localized maximum sharpness value, the log of which produces a dimensionless blur value.

A device includes a respirometer 3 and an automatic apnea/hypopnea analyzer 2. The respirometer 3 includes an airflow-signal recording unit 13 that is connected to a thermistor respiratory flow meter 5 or a nasal-pressure type flow meter 6 detecting the airflow waveform signals, converts to digital data the airflow waveform signals obtained by the thermistor respiratory flow meter 5 or the nasal-pressure type flow meter 6, and stores the converted data as measured respiratory flow values. The automatic apnea/hypopnea analyzer 2 obtains power spectra in the breathing frequency band from the measured respiratory flow values, calculates logarithmic time-series data from the power spectra, smoothes the data, and then detects transitory drops (flow power dips) in the smoothed data, to automatically detect apnea/hypopnea.

Decoder for low-density parity check convolutional codes. In at least some embodiments, a decoder (200) for arbitrary length blocks of low-density, parity-check codes includes a plurality of interconnected processors (202), which further include a plurality of interconnected nodes. A memory can be interconnected with the nodes to store intermediate log likelihood ratio (LLR) values based on channel LLR values. Thus, LLR values having successively improved accuracy relative to the channel LLR values can be output from each processor, and eventually used to decision information bits. In some embodiments, the memory is a random access memory (RAM) device that is adapted to store the intermediate LLR values in a circular buffer. Additionally, a storage device such as a read-only memory (ROM) device can be used to generate a predetermined plurality of addresses for reading and writing LLR values.

Dislosed is an apparatus and method for erasure detection and soft-decision decoding in a cellular system receiver. In the present invention, a complex channel gain and a noise variance of a received symbol from a cellular system transmitter are estimated, and an erasure symbol (a symbol with high power interference) is detected by comparing a threshold value according to the estimated complex channel gain or noise variance with power of the received symbol. Log-likelihood ratios corresponding to constituent bits of the detected erasure symbols are allowed to be 0, and the log-likelihood ratios of the constituent bits of the other received symbols are calculated. A soft-decision decoding operation of received bits is performed with the calculated log-likelihood ratios.

Disclosed are a demand forecasting method, a demand forecasting program, and a demand forecasting system that can adequately forecast a demand for a commodity. A management computer calculates a trend function with respect to demand values. The management computer then calculates the ratio of the demand values to a trend curve, and calculates a logarithm of the ratio. Then, the management computer approximates the calculated natural logarithm curve with a periodic model, and converts the logarithmic scale of the acquired periodic model to a normal scale. Demand forecasting is performed using a demand forecasting model acquired by combining a periodic change model of the normal scale with a trend curve (trend function).

A method according to the invention separates multiple audio signals recorded as a mixed signal via a single channel. The mixed signal is A/D converted and sampled. A sliding window is applied to the samples to obtain frames. The logarithms of the power spectra of the frames are determined. From the spectra, the a posteriori probabilities of pairs of spectra are determined. The probabilities are used to obtain Fourier spectra for each individual signal in each frame. The invention provides a minimum-mean-squared error metho or a soft mask method for making this determination. The Fourier spectra are inverted to obtain corresponding signals, which are concatenated to recover the individual signals.

The present invention describes a method and apparatus for performing logarithmic arithmetic with real and/or complex numbers represented in a logarithmic format. In one exemplary embodiment, an ALU implements logarithmic arithmetic on complex numbers represented in a logpolar format. According to this embodiment, memory in the ALU stores a look-up table used to determine logarithms of complex numbers, while a processor in the ALU generates an output logarithm based on complex input operands represented in logpolar format using the stored look-up table. In another exemplary embodiment, the ALU performs logarithmic arithmetic on real and complex numbers represented in logarithmic format. In this embodiment, the memory stores two look-up tables, one for determining logarithms of real numbers and one for determining logarithms of complex numbers, while the processor generates an output logarithm based on real or complex input operands represented in logarithmic format using the real or complex look-up tables, respectively.

The invention provides an intersection map matching method based on driver behavior characteristics and a logit model, relating to the field of vehicle positioning and map matching. The method comprises four phases of storing input data, mapping an initial site map, matching a map on a road section and matching an intersection map, so as to finish high-precision and real-time positioning of a vehicle driving track. The intersection map matching method disclosed by the invention solves the problem that the matching precision of the intersection map is low and overcomes the defect of the traditional intersection map matching method, so that the high-precision and real-time positioning of an intersection is realized.

The invention discloses a model method of a traffic prediction. The method comprises the following steps of: (1) establishing a road traffic basic facility geographical information database, storing and updating traffic planning schemes, and establishing intersection delay determination; (2) establishing an urban planning construction geographical information database, storing regulatory detailed plans of each area, and utilizing an original unit method to construct traffic generation cases; (3) calling data of the road traffic basic facility geographical information database and the urban planning construction geographical information database, and utilizing a gravity model method to predict parallel distributed matrixes; (4) constructing a multi-element logit model, and determining the probability distribution for a single person selecting a travelling mode in specific travelling information; and (5) utilizing a capacity-limited multi-path distribution method to carry out traffic distribution. The modeling purpose, method to realization are overall planed, modeling parameter selection, basic data sources and calculation program design are coordinated, and the convenient application and operation of a traffic prediction model system are ensured.

In one embodiment, an LDPC decoder has a controller and an extrinsic log-likelihood (LLR) value generator. The extrinsic LLR value generator is selectively configurable to operate in either (i) a non-averaging mode that updates extrinsic LLR values without averaging or (ii) an averaging mode that updates extrinsic LLR values using averaging. Initially, the extrinsic LLR value generator is configured to generate non-averaged extrinsic LLR values, and the decoder attempts to recover an LDPC-encoded codeword using the non-averaged extrinsic LLR values. If the decoder is unable to recover the correct codeword, then (i) the controller selects the averaging mode, (ii) the extrinsic LLR value generator is configured to generate average extrinsic LLR values, and (iii) the decoder attempts to recover the correct codeword using the average extrinsic LLR values. Averaging the extrinsic LLR values may slow down the propagation of erroneous messages that lead the decoder to convergence on trapping sets.