Microsoft Technology Licensing, LLC

Jian-Lai Zhou

Jian Wang

Dongmei Zhang

Lei Huang

Yu Zou

Xiaohui Hou

Mingqing Xu

Peng Liu

Ye Tian

Eric I-Chao Chang

Xianfang Wang

Yue Li

Zhouchen Lin

Jiang Wu

Wenli Zhu

Hongyun Yang

Frank Kao-Ping K Soong

Zheng Chen

Alejandro Acero

Xuedong D Huang

Li Deng

Hua Li

Yong Zhao

Min Chu

Hagai Thomas Attias

Frank Torsten Bernd Seide

Frank Kao-Ping Soong

Lijuan Wang

Asela J R Gunawardana

Frank Kao-ping Soong

Dongmei Zhang, Legal Representative

Measurement of Kullback-Leibler Divergence (KLD) between hidden Markov models (HMM) of acoustic units utilizes an unscented transform to approximate KLD between Gaussian mixtures. Dynamic programming equalizes the number of states between HMMs having a different number of states, while the total KLD of the HMMs is obtained by summing individual KLDs calculated by state pair by state pair comparisons.

Calculating cost measures between HMM acoustic models

A mechanism for recognizing and inputting handwritten mathematical expressions into a computer by providing part of a multi-path framework is described. The part of the multi-path framework includes a subordinate sub-expression analysis component. A method for analyzing a handwritten mathematical expression for a subordinate sub-expression includes identifying sub-expressions based on dominant symbols and determining a character for potential dominant symbols based upon sub-expression information. A determination may be made whether an expression structure candidate is valid and valid expression structure candidates may be stored in a parse tree.

Analyzing subordinate sub-expressions in expression recognition

Speech recognition such as command and control speech recognition generally use a context free grammar to constrain the decoding process. Word or subword background model are constructed to repopulate dynamic hypothesis space, especially when word spareness is at issue. The background models can be later used in speech recognition. During speech recognition, background and conventional context free grammar decoding are used to measure confidence. The discussion above is merely provided for general background information and is not intended to be used as an aid in determining the scope of the claimed subject matter.

Subword unit posterior probability for measuring confidence

Multiple input modalities are selectively used by a user or process to prune a word graph. Pruning initiates rescoring in order to generate a new word graph with a revised best path.

Common word graph based multimodal input

A reliable full covariance matrix estimation algorithm for pattern unit's state output distribution in pattern recognition system is discussed. An intermediate hierarchical tree structure is built to relate models for product units. Full covariance matrices of pattern unit's state output distribution are estimated based on all the related nodes in the tree.

Covariance estimation for pattern recognition

An opinion system infers the opinion of a sentence of a product review based on a probability that the sentence contains certain sequences of parts of speech that are commonly used to express an opinion as indicated by the training data and the probabilities of the training data. When provided with the sentence, the opinion system identifies possible sequences of parts of speech of the sentence that are commonly used to express an opinion and the probability that the sequence is the correct sequence for the sentence. For each sequence, the opinion system then retrieves a probability derived from the training data that the sequence contains an opinion word that expresses an opinion. The opinion system then retrieves a probability from the training data that the opinion words of the sentence are used to express an opinion. The opinion system then combines the probabilities to generate an overall probability that the sentence with that sequence expresses an opinion.

Inferring opinions based on learned probabilities

A mechanism for recognizing and inputting handwritten mathematical expressions into a computer by providing part of a multi-path framework is described. The part of the multi-path framework includes a subscript/superscript analysis and character determination component that is designed to identify subscript and superscript elements. A method for analyzing a handwritten mathematical expression includes receiving a symbols corresponding to handwritten mathematical expression input strokes, identifying subscript and/or superscript structures, and determining a character for each symbol of the set. A graph of vertexes and edges may be created based upon the set of symbols and the graph may be searched to determine optimized candidates.

Analyzing scripts and determining characters in expression recognition

A mechanism for recognizing and inputting handwritten mathematical expressions into a computer by providing a multi-path framework is described. The framework may include symbol grouping and recognition, tabular structure analysis, subordinate sub-expression analysis, subscript/superscript analysis and character determination, and semantic structure analysis components. A method for recognizing a handwritten mathematical expression includes receiving a plurality of input strokes corresponding to a handwritten mathematical expression and providing a candidate list of recognized candidate expressions based upon the input strokes. Input strokes are grouped into symbols, tabular structures are determined, dominant symbol candidates and subordinate symbols are determined, and subscript and superscript structures are determined.

Mathematical expression recognition

A mechanism for recognizing and inputting handwritten mathematical expressions into a computer by providing a part of a multi-path framework is described. The part of the multi-path framework includes a symbol grouping and recognition component that is designed to group input strokes that correspond to a handwritten mathematical expression into a symbol and to recognize the symbol based upon information associated with the grouped input strokes. A method for grouping and recognizing symbols of a handwritten mathematical expression includes receiving a plurality of input strokes corresponding to a handwritten mathematical expression, grouping the plurality of input strokes into symbols, recognizing the symbols based upon information, such as shape and time series information, associated with the grouped input strokes. Intra-group and inter-group information associated with the plurality of input strokes may be utilized to group the input strokes.

Symbol grouping and recognition in expression recognition

Systems, methods, and computer-readable media for processing electronic ink receive an electronic ink input; convert the electronic ink input to a first machine-generated object using handwriting recognition; display the first machine-generated object on a display; receive speech input; convert the speech input to a second machine-generated object using speech recognition; generate a list of machine-generated objects based on the electronic ink input, the list including the first machine-generated object and alternative machine-generated objects and functioning as a dictionary for converting the speech input; and replace the first machine-generated object with the second machine-generated object. A user may confirm that the second machine-generated object should replace the first machine-generated object. The systems and methods may generate a list of alternative machine-generated object candidates to the first machine-generated object based on handwriting recognition of the electronic ink input alone or in combination with a statistical language model.

Multi-modal handwriting recognition correction

Input is received from at least two different input sources. Information from these sources are combined together to provide a result. In a particular example, input from one source corresponds to potential recognition candidates, and input from another source corresponds to other potential candidates. These candidates are combined together to select a result.

Combined input processing for a computing device

A method and apparatus are provided for refining segmental boundaries in speech waveforms. Contextual acoustic feature similarities are used as a basis for clustering adjacent phoneme speech units, where each adjacent pair phoneme speech units include a segmental boundary. A refining model is trained for each cluster and used to refine boundaries of contextual phoneme speech units forming the clusters.

Refining of segmental boundaries in speech waveforms using contextual-dependent models

A speech signal is decoded by determining a production-related value for a current state based on an optimal production-related value at the end of a preceding state, the optimal production-related value being selected from a set of continuous values. The production-related value is used to determine a likelihood of a phone being represented by a set of observation vectors that are aligned with a path between the preceding state and the current state. The likelihood of the phone is combined with a score from the preceding state to determine a score for the current state, the score from the preceding state being associated with a discrete class of production-related values wherein the class matches the class of the optimal production-related value.

Method of speech recognition using time-dependent interpolation and hidden dynamic value classes

A computationally efficient and robust pitch detection and tracking system and related methods are presented. According to certain exemplary implementations a method is presented comprising identifying an initial set of pitch period candidates using a first estimation algorithm, filtering the initial set of candidates and passing the filtered candidates through a second, more accurate pitch estimation algorithm to generate a final set of pitch period candidates from which the most likely pitch value is selected.

Growing community of inventors

Beijing, China

Jian-Lai Zhou