Nortel Networks Corporation

Northern Telecom Limited

(crim) Centre De Recherche Informatique De Montreal

Bell Canada

Vishwa N Gupta

Matthew Lennig

Michael G Sabourin

Peter R Stubley

Serge Robillard

Gregory J Bielby

Jean-Guy Dahan

Hans A Wasmeier

Paul Boucher

Gilles Boulianne

R Douglas Sharp

Lauren C Hodgson

Christopher K Toulson

Real Tremblay

Paul Mermelstein

Jean-Francois Crespo

Waleed Fakhr

Andre Gillet

David Bruce Peters

Patrick J Kenny

Patrick Cardnal

Jianxiong Wu

Parisa Darvish Zadeh Varcheie

A method for performing audio copy detection, comprising, providing a query audio data, the query audio data having a succession of frames and also providing a plurality of test audio data units, each test audio data unit including a succession of frames. For each test audio data unit the method generates a test fingerprint set. The generation of the test fingerprint test including computing similarity measurements between at least one frame of the test audio data and a plurality of frames of the query audio data. A test audio data unit is then selected as a match for the query audio data at least in part on the basis of the fingerprint sets.

Content based audio copy detection

A method to detect video copying based on content. The method comprises providing a set of reference data elements derived from a set of reference video frames in a reference video stream; providing a set of query data elements derived from a set of query video frames in a query video stream, each of the query data elements having a corresponding query data element identifier; associating with each of the reference data elements a fingerprint selected from among the query data element identifiers; and determining a similarity measure for the query video stream relative to the reference video stream by a comparison of the query data element identifiers to the fingerprints.

Content-based video copy detection

Occurrences of one or more keywords in audio data are identified using a speech recognizer employing a language model to derive a transcript of the keywords. The transcript is converted into a phoneme sequence. The phonemes of the phoneme sequence are mapped to the audio data to derive a time-aligned phoneme sequence that is searched for occurrences of keyword phoneme sequences corresponding to the phonemes of the keywords. Searching includes computing a confusion matrix. The language model used by the speech recognizer is adapted to keywords by increasing the likelihoods of the keywords in the language model. For each potential occurrences keywords detected, a corresponding subset of the audio data may be played back to an operator to confirm whether the potential occurrences correspond to actual occurrences of the keywords.

Identifying keyword occurrences in audio data

A data communication terminal, such as a cellular telephone, capable of synthesizing speech. The data communication terminal can establish a communication session with a base station over a transmission facility implementing a voice channel and a data channel. The data communication terminal includes a speech synthesizer engine that receives from the remote entity a signal transmitted over the data channel and that conveys the vocal tract characteristics of the message to be delivered as a spoken announcement. The base station generates the signal containing the vocal tract characteristics from a text-based signal of the message to be synthesized. The invention also extends to a base station that can convert the text based message to be synthesized into a signal containing vocal tract characteristics and that sends the signal to a remote terminal over the data channel.

Method and apparatus for performing text to speech synthesis

A method and an apparatus for adding a new entry to a speech recognition dictionary, and more particularly a system and method for generating transcriptions from multiple utterances and a textual representation of a given word is provided. The method and apparatus combine acoustic information obtained from each of the training utterances to generate the most likely transcriptions. It also provides a method and apparatus for transcribing multiple utterances into a single transcription that can be added to a speech recognition dictionary. In a first step, a graph representative of possible pronunciation of the word or combination of words is generated on the basis of the textual representation of the word and a letter to phoneme set of rules. Phonemes perceptually close to the phonemes in the transcriptions obtained using the letter to phoneme rules are added on parallel branches in the generated graph. Following this, each branch in the graph is replaced with its allophones and the complete graph is scored using training utterances to obtain the most likely allophone sequence on the basis of the multiple utterances of the word.

Method and apparatus for obtaining a transcription of phrases through text and spoken utterances

The invention relates to an automated directory assistance system that utilizes a priori advisor for predicting the most likely requested locality. The automated directory assistance system includes a speech recognition dictionary containing a plurality of orthographies, each orthography corresponding to a locality name in which a subscriber whose telephone number is sought by the user of the automated directory assistance system may be residing. Upon reception of the spoken utterance, the system performs a first pass search scores on the basis of acoustics characteristics of the orthographies in the speech recognition dictionary, each orthography having a certain likelihood of being a match to the spoken utterances. The orthographies are then weighed on the basis of information indicative of the geographical location of the user. A final re-scoring operation may then be performed on the top N candidates in the weighed list. This system enables to improve recognition accuracy by combining the acoustical match search with a probabilistic bias derived from statistical information on calling patterns in the population.

Automated directory assistance system utilizing priori advisor for

Comparing a series of observations representing unknown speech, to stored models representing known speech, the series of observations being divided into at least two blocks each comprising two or more of the observations, is carried out in an order which makes better use of memory. First, the observations in one of the blocks are compared (31), to a subset comprising one or more of the models, to determine a likelihood of a match to each of the one or more models. This step is repeated (33) for models other than those in the subset; and the whole process is repeated (34) for each block.

Method and apparatus for speech recognition

The invention relates to a method and an apparatus for recognising speech, more particularly to a speech recognition system and method utilising a speech recognition dictionary supplemented by a lexicon containing frequently occurring word sequences (orthographies). In typical speech recognition systems, the process of speech recognition consists of scanning the vocabulary database or dictionary by using a fast match algorithm to find the top N candidates that potentially match the input speech. In a second pass the N candidates are re-scored using more precise likelihood computations. The novel method comprises the introduction of a step in the search stage that consists of forcing the insertion in the list of N candidates entries selected from a lexicon containing frequently used orthographies to increase the probability of occurrence of certain text combinations.

Method and apparatus for performing speech recognition utilizing a

Systems and methods consistent with the present invention determine whether to accept one of a plurality of intermediate recognition results output by a speech recognition system as a final recognition result. The system first combines a plurality of speech rejection features into a feature function in which weights are assigned to each rejection feature in accordance with a recognition accuracy of each rejection feature. Feature values are then calculated for each of the rejection features using the plurality of intermediate recognition results. The system next computes the feature function according to the calculated feature values to determine a rejection decision value. Finally, one of the plurality of intermediate recognition results is accepted as the final recognition result according to the rejection decision value.

Speech recognition rejection method using generalized additive models

The invention relates to a method and apparatus for automatically generating probability tables (histograms) for a speech recognition dictionary for use in a speech recognition system from a listing containing call records. The method comprises the step of generating histograms representing the probabilities of localities being requested based on the calling pattern collected on the field by either a human operator or via an automatic process during calls. The method is particularly useful for generating an a priori advisor for a speech recognition dictionary used in an automated system for locality recognition.

Method and apparatus for generating an a priori advisor for a speech

The invention relates to an automated directory assistance system that utilizes a heuristics model for predicting the most likely requested number. Each orthography link in the speech recognition dictionary pointing toward an entry in the white pages is associated with a data structure that provides a probability value of that link pointing toward the telephone number intended by the user on the basis of locality information specified by the user of the automated directory assistance system.

Automated directory assistance system utilizing a heuristics model for

The invention relates to a method and an apparatus for adding a new entry to a speech recognition dictionary, more particularly to a system and method for generating transcriptions from multiple utterances of a given word. The novel method and apparatus automatically transcribes several training utterances into transcriptions without knowledge of the orthography of the word being added. It also provides a method and apparatus for transcribing multiple utterances into a single transcription that can be added to a speech recognition dictionary. In a first step, each utterance is analyzed individually to get their respective acoustic characteristics. Following this, these characteristics are combined to generate a set of the most likely transcriptions using the acoustic information obtained from each of the training utterances.

Method and apparatus for obtaining transcriptions from multiple training

The invention relates to a method and apparatus for grouping orthographies in a speech recognition dictionary to reduce false rejection. In a typical speech recognition system, the process of speech recognition consists of scanning the vocabulary database by using a fast match algorithm to find the top N orthography groups. In a second pass the orthographies in the selected groups are re-scored using more precise likelihoods. The top orthographies in the top two groups are then processed by a rejection algorithm to find if they are sufficiently distinct from one another. In the affirmative the top choice candidate is accepted, otherwise the entire procedure is terminated. The novel method comprises the steps of grouping confusable orthographies together to reduce the possibility of false rejection.

Method and apparatus for reducing false rejection in a speech

The invention relates to a method and apparatus for automatically generating a speech recognition vocabulary for a speech recognition system from a listing that contains a number of entries, each entry containing a multi-word identification data that distinguishes that entry from other entries in the list. The method comprises the steps of creating for each entry in the listing a plurality of orthographies in the speech recognition vocabulary that are formed by combining selected words from the entry. The words combination is effected by applying a heuristics model that mimics the way users formulate requests to the automated directory assistance system. The method is particularly useful for generating speech recognition vocabularies for automated directory assistance systems.

Growing community of inventors

Brossard, Canada

Vishwa N Gupta