专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

31. 发明公开

EP0960417A2 METHOD OF DETERMINING MODEL-SPECIFIC FACTORS FOR PATTERN RECOGNITION, IN PARTICULAR FOR SPEECH PATTERNS 有权
标题翻译：程序为MUSTEREKENNUNG MODEL-特有的因素，特别是对于语音样本
公开(公告)号：EP0960417A2
公开(公告)日：1999-12-01
申请号：EP98957052.0
申请日：1998-12-11
申请人： Koninklijke Philips Electronics N.V. , Philips Patentverwaltung GmbH
发明人： BEYERLEIN, Peter
IPC分类号： G10L15 , G06F17
CPC分类号： G10L15/183 , G06F17/17 , G10L15/142
摘要： A method for recognizing a pattern that comprises a set of physical stimuli, said method comprising the steps of: providing a set of training observations and through applying a plurality of association models ascertaining various measuring values pj(k ¦ x), j=1... M, that each pertain to assigning a particular training observation to one or more associated pattern classes; setting up a log/linear association distribution by combining all association models of the plurality according to respective weight factors, and joining thereto a normalization quantity to produce a compound association distribution; optimizing said weight factors for thereby minimizing a detected error rate of the actual assigning to said compound distribution; recognizing target observations representing a target pattern with the help of said compound distribution.

32. 发明公开

EP0924688A2 Markov model discriminator using negative examples 审中-公开
标题翻译： Markov-Modell- Diskriminator unter Verwendung von negativen Beispielen
公开(公告)号：EP0924688A2
公开(公告)日：1999-06-23
申请号：EP98115193.9
申请日：1998-08-12
申请人： MITSUBISHI DENKI KABUSHIKI KAISHA
发明人： Brand, Matthew E.
IPC分类号： G10L5/06
CPC分类号： G10L15/142 , G06K9/6297 , G10L15/144
摘要： Negative examples are used along with positive examples to modify a Markov Model resulting in lower error rates in classification tasks as compared with conventionally trained Markov models. The subject system is used for identifying particular traits or characteristics of sequences to permit identification of, for instance, inappropriate web page material, hand signing gestures, audio programm material type, authorship of a text, with the system also being useful in speech recognition, as well as seismic, medical, and industrial monitoring.
摘要翻译：与常规训练的马尔科夫模型相比，负面实例与正例一起用于修改马尔可夫模型，导致分类任务中的错误率更低。主题系统用于识别序列的特定特征或特征，以允许识别例如不适当的网页材料，手签署手势，音频节目材料类型，文本的作者，该系统也可用于语音识别，以及地震，医疗和工业监测。

33. 发明授权

EP0592150B1 Speaker verification 失效转让
标题翻译：说话人确认
公开(公告)号：EP0592150B1
公开(公告)日：1999-05-19
申请号：EP93307765.3
申请日：1993-09-30
申请人： AT&T Corp.
发明人： Juang, Biing-Hwang , Lee, Chin-Hui , Rosenberg, Aaron Edward , Soong, Frank Kao-Ping
IPC分类号： G10L5/06 , G10L7/08 , G10L9/06
CPC分类号： G10L17/00 , G10L15/142 , G10L17/12 , H04M3/382 , H04M3/42204 , H04M15/00 , H04M2201/40

34. 发明授权

EP0813734B1 VERFAHREN ZUR ERKENNUNG MINDESTENS EINES DEFINIERTEN, DURCH HIDDEN-MARKOV-MODELLE MODELLIERTEN MUSTERS IN EINEM ZEITVARIANTEN MESSIGNAL, WELCHES VON MINDESTENS EINEM STÖRSIGNAL ÜBERLAGERT WIRD 失效
标题翻译：检测方法的至少一项所定义建模模式是一种时间的变化测量信号隐马尔可夫模型，其中至少有一个噪音LAYERS
公开(公告)号：EP0813734B1
公开(公告)日：1999-04-28
申请号：EP96902240.9
申请日：1996-02-19
申请人： SIEMENS AKTIENGESELLSCHAFT
发明人： ZÜNKLER, Klaus
IPC分类号： G10L5/06
CPC分类号： G10L15/142 , G10L15/20 , G10L2015/088

35. 发明授权

EP0590925B1 Method of speech modelling and a speech recognizer 失效
标题翻译：语音整形的用于语音识别的方法和装置
公开(公告)号：EP0590925B1
公开(公告)日：1999-04-14
申请号：EP93307664.8
申请日：1993-09-28
申请人： International Business Machines Corporation
发明人： Nishimura, Masafumi , Okochi, Masaaki
IPC分类号： G10L5/06 , G10L7/08 , G10L9/06
CPC分类号： G10L15/142 , G10L15/187 , G10L15/197 , G10L2015/025 , G10L2015/0631
摘要： Provided is a speech recognizer representing various pronunciational transformations efficiently by statistical combinations (N-grams) of a few types of hidden Markov models. Analysis of a word input from a speech input device 1 for its features is made by a feature extractor 4 to obtain a feature vector sequence corresponding to said word or a label sequence by applying a further transformation in a labeler 8. Phonemic hidden Markov models for each speech transformation candidate transformed as a sequence of subwords constituting said word are retained in a parameter table 18, keeping an N-gram relationship (N = an integer greater than or equal to 2) with the speech transformation candidate of other preceding subwords in the word. A recognizer 16 then applies hidden Markov models to each speech transformation candidate in correspondence to the candidate words in a word pronunciation dictionary 13 and on the basis of said N-gram relation, and joins each hidden Markov model for each of these speech transformation candidates in parallel among the subwords to compose a speech model. The recognizer determines the probability of the speech model composed for each candidate word outputting said label sequence or feature vector sequence input as speech and outputs a candidate word corresponding to the speech model of the highest probability to a display 19.

36. 发明公开

EP0847041A3 Method and apparatus for speech recognition performing noise adaptation 失效
标题翻译：用于语音识别噪声适配的方法和装置
公开(公告)号：EP0847041A3
公开(公告)日：1999-02-03
申请号：EP97309678.7
申请日：1997-12-02
申请人： CANON KABUSHIKI KAISHA
发明人： Komori, Yasuhiro , Yamamoto, Hiroki
IPC分类号： G10L5/06
CPC分类号： G10L15/142 , G10L15/144 , G10L15/20
摘要： A speech processing apparatus includes a noise model production device for extracting a noise-speech interval from input speech data and producing a noise model by using the data of the extracted interval. The apparatus also includes a composite distribution production device for dividing the distributions of a speech model into a plurality of groups, producing a composite distribution of each group, and determining the positional relationship of each distribution within each group. In addition, the apparatus includes a memory for storing each composite distribution and the positional relationship of each distribution within the group, and a parallel model combination (PMC) conversion device for PMC-converting each produced composite distribution. Also provided is a noise-adaptive speech model production device for producing a noise-adaptive speech model on the basis of the composite distribution which is PMC-converted by the PMC conversion device and the positional relationship stored by the memory. Further, the apparatus includes an output device for determining and outputting a recognition result and a candidate with their likelihood for the input speech data by using the produced noise-adaptive speech model.

37. 发明授权

EP0535909B1 Speech recognition 失效
标题翻译：语音识别
公开(公告)号：EP0535909B1
公开(公告)日：1998-04-01
申请号：EP92308855.3
申请日：1992-09-29
申请人： International Business Machines Corporation
发明人： Nishimura, Masafumi , Kuroda, Akihiro , Toshioka, Koichi
IPC分类号： G10L5/06
CPC分类号： G09B19/04 , G10L15/142 , G10L15/285

38. 发明公开

EP0813734A1 VERFAHREN ZUR ERKENNUNG MINDESTENS EINES DEFINIERTEN, DURCH HIDDEN-MARKOV-MODELLE MODELLIERTEN MUSTERS IN EINEM ZEITVARIANTEN MESSIGNAL, WELCHES VON MINDESTENS EINEM STÖRSIGNAL ÜBERLAGERT WIRD 失效
标题翻译：检测方法的至少一项所定义建模模式是一种时间的变化测量信号隐马尔可夫模型，其中至少有一个噪音LAYERS
公开(公告)号：EP0813734A1
公开(公告)日：1997-12-29
申请号：EP96902240
申请日：1996-02-19
申请人： SIEMENS AG
发明人： ZUENKLER KLAUS
IPC分类号： G10L15/00 , G10L15/14 , G10L15/20 , G10L5/06
CPC分类号： G10L15/142 , G10L15/20 , G10L2015/088
摘要： The invention concerns a particular process for recognising patterns in test signals, i.e. speech signals or signals obtainable from script-recognition devices. In order to carry out the invention, the hidden Markov models with which the pattern is modelled are extended by a special state which has no emission and transition probability, making the position in time of the pattern sought completely irrelevant in terms of its generation probability. Further, the proposed process has the advantage of being able to absorb new and unexpected interferences without the model having to be specially trained for that purpose. Unlike current processes, there is no need to train for the background models. However, this entails a more complex procedure for recognising the patterns, since the individual paths of the Viterbi algorithm must be standardised in terms of their respective probabilities in relation to the current accumulated probabilities in the path, to allow a comparison to be made. The claimed process has the advantage that only the period of the test signal which also contains the pattern needs to be analysed . This ensures a greater hit probability while also reducing computer processing requirements.

39. 发明公开

EP0674306A3 Signal bias removal for robust telephone speech recognition 失效
标题翻译：鲁棒语音识别在电话信号失真的校正。
公开(公告)号：EP0674306A3
公开(公告)日：1997-09-10
申请号：EP95301747.2
申请日：1995-03-16
申请人： AT&T Corp.
发明人： Juang, Biing-Hwang , Rahim, Mazin G.
IPC分类号： G10L3/00
CPC分类号： G10L15/20 , G10L15/02 , G10L15/142 , G10L21/0232
摘要： A signal bias removal (SBR) method based on the maximum likelihood estimation of the bias for minimizing undesirable effects in speech recognition systems is described. The technique is readily applicable in various architectures including discrete (vector-quantization based), semicontinuous and continuous-density Hidden Markov Model (HMM) systems. For example, the SBR method can be integrated into a discrete density HMM and applied to telephone speech recognition where the contamination due to extraneous signal components is unknown. To enable real-time implementation, a sequential method for the estimation of the bias (SSBR) is disclosed.

40. 发明授权

EP0504927B1 Speech recognition system and method 失效
标题翻译：方法和系统的语音识别
公开(公告)号：EP0504927B1
公开(公告)日：1997-07-16
申请号：EP92104898.9
申请日：1992-03-20
申请人： KABUSHIKI KAISHA TOSHIBA
发明人： Nitta, Tsuneo, c/o Intellectual Property Div.
IPC分类号： G10L5/06
CPC分类号： G09B19/04 , G10L15/142 , G10L15/144 , G10L25/12 , G10L2015/025
摘要： A sound analyzer (11) sound analyzes an input speech signal to obtain feature vectors. A matrix quantizer (13) performs a matrix quantization process between the feature vectors obtained by the sound analyzer and a phonetic segment dictionary prepared in phonetic segment units to obtain a phonetic segment similarity sequence. A PS-phoneme integrating section (14) integrates the phonetic segment similarity sequence into a phonemic feature vector. A HMM recognizer (15) checks the phonemic feature vector using a HMM prepared in certain units, to thereby perform a recognition process.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式