Abstract: A bio-signal is monitored while a speech recognition system is trained to recognize a word or utterance. An utterance is identified for retraining when the bio-signal is above an upper threshold or below a lower threshold while the recognition system is being trained to recognize the utterance.

Abstract: The recognition rate of a speech recognition system is improved by compensating for changes in the user's speech that result from factors such as emotion, anxiety or fatigue. A speech signal derived from a user's utterance, and a bio-signal, which is indicative of the user's emotional state, are provided to a speech recognition system. The bio-signal is used to provide a reference frequency that changes when the user's emotional state changes. An utterance is identified by examining the relative magnitudes of its frequency components and the position of the frequency components relative to the reference frequency.

Abstract: A voice coding communication system in case of applying an adaptive differential PCM system in which a discontinuous transmitter is provided at the transmitting side to reduce its power consumption, transmission is suspended during a silent duration, and a pseudo-noise is generated at the receiving side and is used as the reproduced output during the silent duration. Discomfort is avoided in hearing the reproduced voice which is caused by the difference between the level of the pseudo-noise and the level of a background noise at the transmitting side.

Abstract: The recognition rate of a speech recognition system is improved by compensating for changes in the user's speech that result from factors such as emotion, anxiety or fatigue. A speech signal derived from a user's utterance is modified by a preprocessor and provided to a speech recognition system to improve the recognition rate. The speech signal is modified based on a bio-signal which is indicative of the user's emotional state.

Abstract: A speech recognition memory compression method and apparatus subpartitions probability density function (pdf) space along the hidden Markov model (HMM) index into packets of typically 4 to 8 log-pdf values. Vector quantization techniques are applied using a logarithmic distance metric and a probability weighted logarithmic probability space for the splitting of clusters. Experimental results indicate a significant reduction in memory can be obtained with little increase in overall speech recognition error.

Abstract: In a speaker verification system, a method of compensating for differences in speech samples obtained during registration and those obtained during verification due to the use of different types of microphones is provided by filtering at least one of the samples such that the similarities of the two samples are increased. The filtered sample is used within the speaker verification matching process. A two-way comparison is disclosed in which both a verification speech sample and a reference sample are filtered with nonlinear microphone characteristics such as carbon microphone characteristics. A four-way comparison is also disclosed in which patterns produced from unfiltered verification and reference samples and patterns produced from the filtered verification and reference samples are compared to identify a match. A score is determined for each comparison. The comparison having the best score is used to determine if a match has occurred.

Abstract: At the time of training reference speech, the relationship between durations of each recognition unit is obtained by a duration training circuit and, at the time of recognizing speech, a beginning and an end of input speech is detected by a speech period sensing circuit, and then by using the relationship and the input speech period length, the durations of the recognition units in the input speech are estimated. Next, the reference speech and the input speech are matched by the matching means by using the calculated estimation values in such a manner that the recognition units have a duration close to that of the estimated values.

Abstract: A method and system for reducing perplexity in a speech recognition system based upon determined geographic location. In a mobile speech recognition system which processes input frames of speech against stored templates representing speech, a core library of speech templates is created and stored representing a basic vocabulary of speech. Multiple location-specific libraries of speech templates are also created and stored, each library containing speech templates representing a specialized vocabulary for a specific geographic location. The geographic location of the mobile speech recognition system is then periodically determined utilizing a cellular telephone system, a geopositioning satellite system or other similar systems and a particular one of the location-specific libraries of speech templates is identified for the current location of the system.

Abstract: A speech coding apparatus and method uses classification rules to code an utterance while consuming fewer computing resources. The value of at least one feature of an utterance is measured during each of a series of successive time intervals to produce a series of feature vector signals representing the feature values. The classification rules comprise at least first and second sets of classification rules. The first set of classification rules map each feature vector signal from a set of all possible feature vector signals to exactly one of at least two disjoint subsets of feature vector signals. The second set of classification rules map each feature vector signal in a subset of feature vector signals to exactly one of at least two different classes of prototype vector signals. Each class contains a plurality of prototype vector signals. According to the classification rules, a first feature vector signal is mapped to a first class of prototype vector signals.

Abstract: A quantization process proposes a low data rate for predictor filters of a vocoder with a speech signal broken down into packets having a predetermined number L of frames of constant duration and a weight allocated to each frame according to the average strength of the speech signal in the respective frame. The process involves allocating a predictor filter for each frame and determining the possible configurations for predictor filters having the same number of coefficients and the possible configuration for which the coefficients of a current frame predictor filter are interpolated from the predictor filter coefficients from neighboring frames. Subsequently, a deterministic error is calculated by measuring the distances between the filters in order to form a first stack with a predetermined number of configurations which give the lowest errors.

Abstract: A speech-recognition system for recognizing isolated words includes pre-processing circuitry for performing analog-to-digital conversion and cepstral analysis, and a plurality of neural networks which compute discriminant functions based on polynomial expansions. The system may be implemented using either hardware or software or a combination thereof. The speech wave-form of a spoken word is analyzed and converted into a sequence of data frames. The sequence of frames is partitioned into data blocks, and the data blocks are then broadcast to a plurality of neural networks. Using the data blocks, the neural networks compute polynomial expansions. The output of the neural networks is used to determine the identity of the spoken word. The neural networks utilize a matrix-inversion or alternatively a least-squares estimation training algorithm which does not require repetitive training and which yields a global minimum to each given set of training examples.

Abstract: In an encoding device (100) operable in response to an input speech signal by means of an adaptive transform coding to produce an output encoded speech signal, the input speech signal is partitioned into data blocks by a partition circuit (113). Each of data blocks is decomposed into a plurality of frequency components by a Fourier transformer (114). A spectral envelope calculator (120) estimates intensity of a spectral envelope of the input speech signal. In cooperation with a scalar spectral calculator (115) and a bit assignment determiner (121), a quantizer (116) quantizes or encodes the frequency components with phase information selectively removed from a part of the frequency components on the basis of the intensity of the spectral envelope. In a decoding device, a phase information assignor assigns pseudo-phase information to each of the frequency components from which the phase information is selectively removed.

Abstract: A speech recognition system starts by training hidden Markov models for all triphones, diphones, and phonemes occurring in a small training vocabulary. Hidden Markov models of a target vocabulary are created by concatenating the triphone, diphone, and phoneme models, using triphone models if available, diphone HMMs when triphone models are not available, and phoneme models when neither triphone nor diphone models are available. Utterances from the target vocabulary are recognized by choosing a model with maximum probability of reproducing quantized utterance features.

Abstract: The invention is a method of operating a robot in successive sampling intervals to perform a task, the robot having joints and joint actuators with actuator control loops, by decomposing the task into behavior forces, accelerations, velocities and positions of plural behaviors to be exhibited by the robot simultaneously, computing actuator accelerations of the joint actuators for the current sampling interval from both behavior forces, accelerations velocities and positions of the current sampling interval and actuator velocities and positions of the previous sampling interval, computing actuator velocities and positions of the joint actuators for the current sampling interval from the actuator velocities and positions of the previous sampling interval, and, finally, controlling the actuators in accordance with the actuator accelerations, velocities and positions of the current sampling interval.

Abstract: A concatenator for a first digital frame with a second digital frame, such as the ending and beginning of adjacent diphone strings being concatenated to form speech is based on determining an optimum blend point for the first and second digital frames in response to the magnitudes of samples in the first and second digital frames. The frames are then blended to generate a digital sequence representing a concatenation of the first and second frames with reference to the optimum blend point. The system operates by first computing an extended frame in response to the first digital frame, and then finding a subset of the extended frame with matches the second digital frame using a minimum average magnitude difference function over the samples in the subset. The blend point is the first sample of the matching subset.

Abstract: A speech processing apparatus for obtaining a processed speech which is natural and comfortable for a listener, by refining a gain value assigned for each frequency band in enhancing formants in a power spectrum. The power spectrum, calculated in a frequency analyzing unit, is subject to contrast enhancement in a contrast enhancing unit, and judged as to whether it is a format or not in each frequency band. In a gain value assigning unit, a gain value of 1 is assigned to a formant, and a gain value smaller than 1 is to a frequency other than formant. A threshold value for each frequency band is determined by a threshold value determining unit in accordance with power spectrum of input speech signal, to eliminate the effect of variation in speech level.

Abstract: An audio signal is coded by converting the audio signal into a signal in a frequency domain and effecting a big allocation on the converted audio signal. A masking threshold level for reducing aurally recognized noise due to a masking effect is determined using a signal spectral distribution in a present frame and a signal spectral distribution in a past frame. If the difference between the determined masking threshold level and a masking threshold level in the past frame is equal to or greater than a predetermined level, then a level limited to the difference corresponding to the predetermined level is regarded as a masking threshold level in the present frame.

Abstract: A speech recognizer that selects a command model for a current sound if the best match score for the current sound exceeds its corresponding threshold score. The threshold score is assigned a confidence score based on the best match score and recognition threshold of a prior sound.

Abstract: A computer input system employing an automatic speech recognizer. The system is used for finding names in a data-base. The system reduces the number of letters a speaker needs to enter in order to find a name uniquely in the data-base. The system enables the speaker to enter inputs that identify which words in the name the speaker's letter inputs correspond to. The system builds a set of search parameters incorporating both the user's word identifier inputs and letter inputs. This set can be called an abbreviation because it usually represents a small fraction of the total number of letters in the name the user seeks to find.