Abstract: An apparatus for recognizing predetermined patterns comprises an input for signal data; a pattern matching stage including a plurality of stores, each storing a representation of a pattern to be recognized; a similarity measurer for producing, for each representation, a measure of the similarity between the signal data and that representation; and a classification stage for deciding from the similarity measures which, if any, of the patterns to which the representations correspond is present in the signal data, and for indicating the presence of any such pattern. The classification stage makes its determination based on all the similarity measures after having been trained using information relating to known patterns and the similarity measurer from the pattern matching stage produced by inputting known patterns.

Abstract: In the speech recognition system disclosed herein, an input utterance is submitted to both a large vocabulary isolated word speech recognition module and a small vocabulary continuous speech recognition module. The small vocabulary contains command words which can be combined in sequences to define commands to an application program. The two recognition modules generate respective scores for identified large vocabulary models and for sequences of small vocabulary models. The score provided by the continuous speech recognizer is normalized on the basis of the length of the speech input utterance and an arbitration algorithm selects among the candidates identified by the recognition modules. Without requiring the user to switch modes, text is output if a score from the isolated word recognizer is selected and a command is output if a score from the continuous speech recognizer is selected.

Abstract: A system and method for adaptation of a speaker independent speech recognition system for use by a particular user. The system and method gather acoustic characterization data from a test speaker and compare the data with acoustic characterization data generated for a plurality of training speakers. A match score is computed between the test speaker's acoustic characterization for a particular acoustic subspace and each training speaker's acoustic characterization for the same acoustic subspace. The training speakers are ranked for the subspace according to their scores and a new acoustic model is generated for the test speaker based upon the test speaker's acoustic characterization data and the acoustic characterization data of the closest matching training speakers. The process is repeated for each acoustic subspace.

Abstract: A speech recognition apparatus using a neural network. A neuron-like element according to the present invention has a means for storing a value of the inner condition thereof, a means for updating a value of internal status on the basis of an output from the neuron-like element itself, outputs from other neuron-like elements and an external input, and an output value generating means for converting a value of internal status into an external output. Accordingly, the neuron-like element itself can retain the history of input data. This enables the time series data, such as speech to be processed without providing any special means in the neural network.

Abstract: An adaptive speech recognition and control system and method for controlling various mechanisms and systems in response to spoken instructions and in which spoken commands are effective to direct the system into appropriate memory nodes, and to respective appropriate memory templates corresponding to the voiced command. Spoken commands from any of a group of operators for which the system is trained may be identified, and voice templates are updated as required in response to changes in pronunciation and voice characteristics over time of any of the operators for which the system is trained.

Abstract: In methods and apparatus for distinguishing stationary signals from non-stationary signals, a set of Linear Predictive Coding (LPC) coefficients characterizing spectral properties of the signal for each of a plurality of successive time intervals, including a current time interval, is determined. The LPC coefficients are averaged over a plurality of successive time intervals preceding the current time interval, and a cross-correlation of the LPC coefficients for the current time interval with the averaged LPC coefficients is determined. The signal is declared to be stationary in the current time interval when the cross-correlation exceeds a threshold value, and is declared to be non-stationary in the current time interval when the cross-correlation is less than the threshold value. The methods and apparatus are particularly applicable to detection of transitions between an absence of speech state, characterized by a stationary signal, and a presence-of-speech state characterized by a non-stationary signal.

Abstract: An improved speech recognition system with at least two speech recognition engines which may or may not be identical. Each of the speech recognition engines supply a recognized-text output signal each of which are provided to a text comparator. The comparator compares the recognized-text output signals and accepts or rejects the text based on the degree of agreement between the output signals from each of the engines. Completely different types of speech recognition engines can be employed. Alternatively, identical speech recognition engines could be employed with a slight perturbation added to one of the engines or the input signal of one of the engines. Alternatively, the very same speech recognition engine could be used at two different points in time, with one receiving some type of perturbation.

Abstract: A speech recognition apparatus recognizes utterances of unnatural speech having a higher performance of recognition accuracy with a smaller amount of speech learning data. The speech recognition apparatus includes an acoustic-phonetic variability learn unit, a normal speech model memory, a spectrum smooth-modifier and a speech recognizer. An input speech signal is acoustically analyzed and transformed into a time-series feature vector. The acoustic-phonetic variability learn unit learns an acoustic-phonetic change of spectrum caused by unnatural speech and generates a plurality of acoustic-phonetic variability models. The normal speech model memory stores a normal speech model learned based on normal speech data. The spectrum smooth-modifier modifies the normal speech model based on a plurality of the acoustic-phonetic variability model and generates a plurality of spectrum-modified speech models.

Abstract: To train a speech recognizer, a new voice message (one or a few isolated words), after being spoken by a user, is converted into a token. The token is then compared with a plurality of templates stored in the recognizer and a recognition score is obtained each time. The templates previously stored in the recognizer include templates for previously trained voice messages and one or more previously formed templates of the new voice message. Three tests are applied to the recognition scores to determine if the token and one of the previously formed templates of the new voice message can become paradigm templates, if the new voice message is too close in pronunciation to a voice message the recognizer has been previously trained to recognize, or if the user should repeat the new voice message to form another token. This training procedure provides a certain level of automatic control over the training process of a speaker dependent speech recognizer in an otherwise unsupervised environment.