35 results about "Joint likelihood" patented technology

The claimed subject matter includes a system and method for recognizing mixed speech from a source. The method includes training a first neural network to recognize the speech signal spoken by the speaker with a higher level of a speech characteristic from a mixed speech sample. The method also includes training a second neural network to recognize the speech signal spoken by the speaker with a lower level of the speech characteristic from the mixed speech sample. Additionally, the method includes decoding the mixed speech sample with the first neural network and the second neural network by optimizing the joint likelihood of observing the two speech signals considering the probability that a specific frame is a switching point of the speech characteristic.

The present invention relates to a system and methodology to facilitate machine learning and predictive capabilities in a processing environment. In one aspect of the present invention, a Mutual Information Model is provided to facilitate predictive state determinations in accordance with signal or data analysis, and to mitigate classification error. The model parameters are computed by maximizing a convex combination of the mutual information between hidden states and the observations and the joint likelihood of states and observations in training data. Once the model parameters have been learned, new data can be accurately classified.

A method for reconstructing image component densities of an object includes acquiring multi-spectral x-ray tomographic data, performing a material decomposition of the multi-spectral x-ray tomographic data to generate a plurality of material sinograms, and reconstructing a plurality of material component density images by iteratively optimizing a functional that includes a joint likelihood term of at least two of the material decomposed sinograms. An x-ray tomography imaging system and a non-transitory computer readable medium are also described herein.

A method for determining a semantic concept associated with an audio signal captured using an audio sensor. A data processor is used to automatically analyze the audio signal using a plurality of semantic concept detectors to determine corresponding preliminary semantic concept detection values, each semantic concept detector being adapted to detect a particular semantic concept. The preliminary semantic concept detection values are analyzed using a joint likelihood model based on predetermined pair-wise likelihoods that particular pairs of semantic concepts co-occur to determine updated semantic concept detection values. One or more semantic concepts are determined based on the updated semantic concept detection values. The semantic concept detectors and the joint likelihood model are trained together with a joint training process using training audio signals, at least some of which are known to be associated with a plurality of semantic concepts.

The present disclosure relates to an acoustic position determination system that includes a mobile communication device and at least one base transmitter unit. The mobile communication device is configured to identify a peak in the received signal, and to de-convolve the signal with all codes that are relevant to the area in which the signal is received. A joint likelihood that a potential code is correct is formed by determining a likelihood based on a signal parameter such as signal-to-noise ratio and a likelihood based on time-of arrival information.

The embodiment of the invention provides a prediction method and method based on multi-source heterogeneous data. The prediction method comprises the steps that at least two categories of historical data which can reflect an event result are obtained according to a set type of events; a joint likelihood model of attribute data of the set type of events and the historical data is built, and optimal estimation of the attribute data is determined according to the maximum posteriori principle; parameters of probability distribution related with the joint likelihood model in the attribute data are determined and used as a prediction result of a to-be-predicted event belonging to the set type of events. According to the prediction method and method based on the multi-source heterogeneous data, the hierarchical model is used, data from difference sources are introduced into different data layers, the heterogeneous data are unified in the same joint likelihood model to be analyzed, and therefore more accurate, real-time and stable prediction result can be obtained through effective fusion.

The invention discloses a hyperspectral image super-resolution processing method based on a convolutional network, mainly solves the problem that a hyperspectral image generated in the prior art is low in resolution, and adopts the implementation scheme that firstly, a training sample and a test sample are formed by an acquired low-resolution hyperspectral image and a high-resolution multispectralimage; then constructing a convolutional network composed of an inference sub-network and a generation sub-network; training the convolutional network by using the training sample, and obtaining theconvolutional network with the highest similarity between the approximate distribution and the real distribution by maximizing a joint likelihood function of the low-resolution hyperspectral image andthe high-resolution multispectral image; and finally, inputting the test sample into the trained convolutional network, and carrying out optimization processing on the generated high-resolution hyperspectral image to obtain a final high-resolution hyperspectral image. According to the method, the resolution of generating the high-resolution hyperspectral image is improved by utilizing the deep convolutional neural network, and the method can be used for medical diagnosis, remote sensing, computer vision and monitoring.

The invention provides a positron emission cerenkov-gamma bi-radiation imaging method. The positron emission cerenkov-gamma bi-radiation imaging method comprises the following steps: placing a visible light photon detector and a gamma photon detector, and acquiring a pulse data set; calculating a joint likelihood probability function of a multi-dimensional data sample during each time period; judging that whether the currently received data slot comes from a positron emission event or not; accumulating all the positron emission events; establishing a transfer function of the system for each voxel; and inverting the input of the transfer function. The invention further provides a positron emission cerenkov-gamma bi-radiation imaging device. The positron emission cerenkov-gamma bi-radiation imaging device comprises a proton-rich isotope injection module, a multi-radiation detector module, a multi-case time coincidence module, a system transfer function acquiring module and a nuclide distribution image reconstruction module. With the adoption of the method and the device provided by the invention, the spatial resolution, device sensitivity and imaging signal-to-noise ratio of the positron imaging device can be effectively improved, and the positron emission cerenkov-gamma bi-radiation imaging method is particularly suitable for the application of the positrons in nondestructive testing and biomedical imaging.

A method for generating a probability map for a cell served by a network node is provided. The method includes generating a first probability map (P1) indicating a likelihood of primary beam directions (step 602). Generating the first probability map includes recording a first direction of a first user equipment (UE), the first direction indicating a direction of a first beam (e.g., a first set ofantenna weights (a.k.a., precoding vector)) associated with the first UE when the first UE appears in the cell served by the network node. The method further includes generating a second probabilitymap (P2) indicating a joint likelihood of primary and secondary beam directions (step 604). Generating the second probability map includes recording a second direction of a second beam associated withthe first UE when the first UE switches from the first beam to the second beam, the second direction being recorded in association with the first direction.

The invention discloses a graph learning model based on a reconstructed graph, and belongs to the field of image annotation, and the method comprises the following steps: searching a semantic nearestneighbor of a test image through an improved nearest neighbor algorithm, constructing a similar matrix, carrying out the clustering of the image through a random dot product graph, mining the internalconnection of the image, obtaining a weighted similar matrix, and obtaining a preliminary image annotation result through a graph learning algorithm. The relation between the labels is used for labeling, the co-occurrence imbalance between the labels is considered in the process, a nearest graph theory model is introduced, and the problem of label imbalance is effectively solved. The random dot product image is used for reconstructing a transfer matrix of labels, and the problem of image label coexistence asymmetry is solved. Further, a Naive Bayes nearest neighbor classifier is used to establish a joint likelihood function between the image and the tag. The image labeling model based on the reconstructed image model is provided according to the characteristic of unbalanced classificationof the image labels, and the recall rate of the labels can be effectively increased.

The invention discloses a momentum wheel reliability evaluation method based on information fusion. By analyzing the magnitude of the performance parameter influenced by the external environment, choosing the remaining oil quantity of the lubricating system which is less affected by the external environment as the key performance index, through the analysis of degradation data, the joint likelihood function is constructed according to the properties of the degradation model, which successfully fuses the performance degradation data information with a small amount of sample lifetime information, estimates the parameters of the degradation model by using the maximum likelihood estimation method, and finally evaluates the reliability of the momentum wheel on the basis of the fused information. The invention makes full use of multi-source information in the reliability evaluation of momentum wheel, and improves the robustness of the evaluation result.

The invention discloses a downlink time frequency synchronization method jointly using a synchronization sequence and OFDM cyclic prefixes. The method achieves downlink time frequency synchronizationof a communication system by jointly using the synchronization sequence and multiple OFDM cyclic prefixes and can solve the problem that in a communication scene, due to factors such as large frequency offset and low signal to noise ratio, time frequency synchronization of an OFDM system is difficult. According to the method, one-dimensional time offset searching is conducted according to a time offset likelihood function, and after a time offset estimation value is obtained, a frequency offset estimation value is obtained by conducting one-dimensional frequency offset searching on a frequencyoffset likelihood function or directly using a low-complexity frequency offset estimation method. In the method, in a timing synchronization stage, time offset likelihood function values of multiplesynchronization cycles can be selected to be subjected to incoherent combination to improve the estimation accuracy degree. Compared with a synchronization method that two-dimensional time offset searching is directly conducted according to the time frequency joint likelihood function to obtain time offset and frequency offset, the method has the advantage that the searching complexity is greatlyreduced. The method is not only suitable for an initial synchronization stage of a communication system, but also suitable for a tracking synchronization stage of the communication system.

The invention discloses a multi-sensor joint detection method based on decision-level and signal-level data fusion. The multi-sensor joint detection method comprises the following steps: S1, obtaining observation values of target signals acquired by node sensors; S2, transmitting observation values acquired by the node sensors under different transmission conditions to a fusion center in different data transmission forms; S3, establishing a joint likelihood function according to the prior information; S4, performing maximum likelihood estimation on unknown parameters; S5, calculating a detection statistic based on a generalized likelihood ratio criterion; S6, setting a detection threshold by deducing approximate distribution of detection statistics; S7, comparing the detection statistic with the detection threshold, and outputting a detection result. According to the method, through combined utilization of multi-source and multi-level data, target information in the data can be reserved as much as possible, the problem that the information utilization rate is insufficient due to the fact that different levels of data are difficult to fuse is effectively solved, and meanwhile the detection performance of a multi-sensor system can be further improved by setting the optimal local judgment threshold value.