Abstract: A unified, nonlinear, non-stationary, stochastic model is disclosed for estimating and removing effects of background noise on speech cepstra. Generally stated, the model is a union of dynamic system equations for speech and noise, and a model describing how speech and noise are mixed. Embodiments also pertain to related methods for enhancement.

Abstract: The present invention obtains a set of text segments from a cluster of different articles written about a common event. The set of text segments is then subjected to textual alignment techniques to identify paraphrases from the text segments in the text. The invention can also be used to generate paraphrases.

Abstract: A system that can receive information from a plurality of simultaneous remote conferences and scan the information for a key indicia. Upon determining that a key indicia is present in the information, the system is operative to alert a user.

Abstract: Methods and systems for sample rate conversion convert a sampled signal to a higher data rate signal. Conversion pulses are received, having a conversion rate that is higher than the sample rate of the sampled signal. Sample points are then reconstructed from the sampled signal, in real time, on either side of a conversion pulse. An interpolation is performed between the reconstructed sample points, at the time of the conversion pulse. The interpolation results are outputted in real time. The process is repeated for additional conversion pulses. The outputted interpolated amplitudes form the higher data rate signal having a data rate equal to the conversion rate. Sample rate conversion is thus performed in real time according to the higher data rate clock, rather than with fixed ratios. As a result, when the higher data rate clock is affected by, for example, jitter or other frequency variations, the higher data rate samples immediately track the lower data rate samples.

Abstract: During a learning phase, a speech recognition device generates parameters of an acceptance voice model relating to a voice segment spoken by an authorized speaker and a rejection voice model. It uses normalization parameters to normalize a speaker verification score depending on the likelihood ratio of a voice segment to be tested and the acceptance model and rejection model. The speaker obtains access to a service application only if the normalized score is above a threshold. According to the invention, a module updates the normalization parameters as a function of the verification score on each voice segment test only if the normalized score is above a second threshold.

Abstract: A method and apparatus are provided for storing parameters of a deleted interpolation language model as parameters of a backoff language model. In particular, the parameters of the deleted interpolation language model are stored in the standard ARPA format. Under one embodiment, the deleted interpolation language model parameters are formed using fractional counts.

Abstract: In the invention, words in f-structure are represented as illustrations that can be understood by any persons regardless of what languages they use, whereby it is made possible to make the f-structure completely language-independent representation. Accordingly, two translation systems of a translation system from one language L into f-structure using illustration representation and a translation system from f-structure using illustration representation into language L are simply constructed, whereby communication support among persons using every language can be provided.

Abstract: A computer-readable medium stores a first lexicon data structure for lexicon words. The first data structure includes a host form variant field containing a host form variant such as a clitic host form variant, a host form field containing the host form of the host form variant (only present if the forms differ) such as a clitic host verbal form, and a verification field indicative of whether the host form variant is a valid word. The first data structure also includes a segment association field containing data or segmentation bits associating the host form variant with certain types of attachment entries in the lexicon, which also contain data or segmentation bits, to define valid combinations between the host form variant and at least one of the attachment entries in the lexicon. A second lexicon data structure for each of the attachment entries in the lexicon is also stored.

Abstract: The invention disclose: Chinese spelling scheme, Chinese alphabetic writing and Phonetic Symbols scheme. The Chinese spelling scheme which is used English letters to mark the tones, it is reduced to 5 sound symbols of the monosyllabic words from 6 spelling letters of other schemes at most. Between the syllables is clearly demarcated. It can be used as marks in sound and tone for Chinese characters, and also can form an alphabetic writing independently which follows the law of international languages. It's highly integrated with word, sound, and code which can be displayed each other and converted each other. It can be translated directly with the Chinese-language sentence and foreign languages. The Phonetic Symbols do not go beyond the range of 26 letters and symbols of common used English keyboard, which can not only be used to mark in English and other languages, but also can be used as a phonetic symbol written language independently.

Abstract: A method of modifying the operation of the encoder function and/or the decoder function of a perceptual coding system in accordance with supplemental information, such as a watermark, so that the supplemental information may be detectable in the output of the decoder function. One or more parameters are modulated in the encoder function and/or the decoder function in response to the supplemental information.

Abstract: A computer method, system and code, for representing a natural-language document in a vector form suitable for text manipulation operations are disclosed. The method involves determining (a) for each of a plurality of terms selected from one of (i) non-generic words in the document, (ii) proximately arranged word groups in the document, and (iii) a combination of (i) and (ii), a selectivity value of the term related to the frequency of occurrence of that term in a library of texts in one field, relative to the frequency of occurrence of the same term in one or more other libraries of texts in one or more other fields, respectively. The document is represented as a vector of terms, where the coefficient assigned to each term includes a function of the selectivity value determined for that term.

Abstract: A pitch adaptive circuit (200) includes an equalizer control circuit (206) that evaluates the pitch of the speech signals that are being processed and depending on the pitch information, the equalizer control circuit (206) selects an equalizer (208, 210) to shape the decoded speech signals. By selecting the best equalizer (208 or 210) to use based on the pitch information, improvements in audio quality are provided automatically without user intervention.

Abstract: A method and apparatus converts input data such as short notes into a global text realization to provide semantically-coherent grammatical text. In various exemplary embodiments, an individual inputs short notes into a computer system, the computer system associates local text realizations with the short notes. Subsequently, the user may select the appropriate local text realizations, which may be converted to semantic representations and to semantically coherent grammatical text or a global text realization.

Abstract: A method for pre-processing speech, in particular for recognizing speech, including receiving a speech signal, separating a spectrum of said speech signal into a number of predetermined frequency sub-bands, analyzing said speech signal within each of said frequency sub-bands, generating respective band-dependent acoustic feature data for each of said respective frequency sub-bands, deriving band-dependent likelihoods for occurrences of speech elements or within said speech signal based on said band-dependent acoustic feature data, analyzing said speech signal within said spectrum, generating full-band acoustic feature data, which are at least in part representative for said speech signal with respect to said spectrum, deriving a full-band likelihood for occurrences of speech elements or of sequences thereof within said speech signal based on said full-band acoustic feature data, deriving an overall likelihood for occurrences of speech elements within said speech signal based on said band-dependent likelihoods

Abstract: The invention includes devices and methods for providing phrase-based translation. Devices of the invention have translation modules containing a set of phrases and their translations wherein the set of phrases is selected based on their suitability in the context in which the translation device is used. The methods of the invention determine the context in which a translation device is used and provide sets of phrases and their translations that are suitable in the context. The invention further includes devices containing software for executing the methods of the invention.

Abstract: A method and system for voice enabled text entry. The method includes receiving a first user-entry with a communications device and creating a transient grammar set as a result of receiving the first user-entry, where the transient grammar set is based on the first user-entry. The method further includes providing an indication that the transient grammar set is created, the indication being provided using the communications device. The method still further includes receiving a second user-entry with the communications device and responsively selecting a text string from the transient grammar set, where the text string corresponds with the second user-entry.

Abstract: The present invention provides a system and method for voice based network management. A network management system uses voice based scripts that allow network management operations to be performed over a phone. In one embodiment, the voice based scripts use a Voice Extensible Markup Language (VXML). However, the voice based script can be any form of communication that allows a user to conduct network management operations using Dual Tone Multi-Frequency (DTMF) signaling or speech commands.

Abstract: A distribution unit 1 divides PCM audio data in plurality of divided data. Each divided data has overlapping sections overlapping with previous and following divided data. An MP3 encoding unit 2 encodes each divided data individually into MP3 data. A filtering process at the time of encoding is executed by using the overlapping sections as similar to before the division of the data. An analyzing unit 3 analyzes the overlapping sections of each divided data encoded into the MP3 data and searches a frame where main data (bit storage values) are not overlapping. A combiner unit 4 combines the adjoining divided data at the searched frames as combining frames. Although the audio data is divided and the divided data are encoded to MP3 data by parallel processes, a continuity of the data and compressing quality can be maintained.

Abstract: Updating a semantic knowledge base derived from sentences obtained from users during the process of creating their documents could improve the comprehensiveness of a semantic knowledge base, and permit the timely acquisition and dissemination of new knowledge. The present invention attains faster updating of semantic knowledge by transmitting unknown sentences collected from users, distributed over a computer network. The user inputs a sentence while creating a document in a particular knowledge domain. If a sentence is not part of the knowledge domain, it is transmitted over the network to the central knowledge source, where it is analyzed and the corresponding sentence and semantic proposition(s) added to the knowledge base as needed. The invention is able to disseminate the knowledge, i.e. propositions, which belong to the knowledge domain, to clients using either a “push” or “pull” methodology.

Abstract: A method and computer-readable medium are provided that perform a series of steps associated with machine translation. These steps include using a first text in a first language and a second text in a second language, to produce an association list where words in the first language are associated with words in the second language. A first syntactic structure for a sentence from the first text is aligned with a second syntactic structure for a sentence in the second text based on the association list without referring to a bilingual dictionary of content words. The association list is also used during translations. Specifically, a word in the first language is translated into a word in the second language based on an entry in the association list without referring to a bilingual dictionary that contains content words. Thus, training and translation are performed without the need for a bilingual dictionary of content words.