Abstract: This invention relates to a method of searching spoken audio data for one or more search terms comprising performing a phonetic search of the audio data to identify likely matches to a search term and producing textual data corresponding to a portion of the spoken audio data including a likely match. An embodiment of the method comprises the steps of taking phonetic index data corresponding to the spoken audio data, searching the phonetic index data for likely matches to the search term, wherein when a likely match is detected a portion of the spoken audio data or phonetic index data is selected which includes the likely match and said selected portion of the spoken audio data or phonetic index data is processed using a large vocabulary speech recognizer. The large vocabulary speech recognizer may derive textual data which can be used for further processing or may be used to present a transcript to a user.

Abstract: This invention relates to a method of searching spoken audio data for one or more search terms comprising performing a phonetic search of the audio data to identify likely matches to a search term and producing textual data corresponding to a portion of the spoken audio data including a likely match. An embodiment of the method comprises the steps of taking phonetic index data corresponding to the spoken audio data, searching the phonetic index data for likely matches to the search term, wherein when a likely match is detected a portion of the spoken audio data or phonetic index data is selected which includes the likely match and said selected portion of the spoken audio data or phonetic index data is processed using a large vocabulary speech recogniser. The large vocabulary speech recogniser may derive textual data which can be used for further processing or may be used to present a transcript to a user.

Abstract: A method of recognizing a radar target comprises producing a sequence of Doppler spectra of radar returns form a scene and producing therefrom a sequence of Doppler feature vectors for a target in the scene. Hidden Markov modelling (HMM) is then used to identify the sequence of Doppler feature vectors as indicating a member of a particular class of targets. HMM is used to identify the sequence of Doppler feature vectors by assigning to each feature vector an occurrence probability by selecting a probability distribution or state from a set thereof associated with a class of targets, multiplying the occurrence probabilities together to obtain an overall probability, repeating for other probability distributions in the set to determine a combination of probability distributions giving highest overall probability for that class of target, then repeating for at least one other class of targets and selecting the target class as being that which yields the highest overall occurrence probability.

Abstract: A recognition system (10) incorporates a filterbank analyser (16) producing successive data vectors of energy values for twenty-six frequency intervals in a speech signal. A unit (18) compensates for spectral distortion in each vector. Compensated vectors undergo a transformation into feature vectors with twelve dimensions and are matched with hidden Markov model states in a computer (24). Each matched model state has a mean value which is an estimate of the speech feature vector. A match inverter (28) produces an estimate of the speech data vector in frequency space by a pseudo-inverse transformation. It includes information which will be lost in a later transformation to frequency space. The estimated data vector is compared with its associated speech signal data vector, and infinite impulse response filters (44) average their difference with others. Averaged difference vectors so produced are used by the unit (18) in compensation of speech signal data vectors.

Abstract: A method of recognising a radar target comprises producing a sequence of Doppler spectra of radar returns form a scene and producing therefrom a sequence of Doppler feature vectors for a target in the scene. Hidden Markov modelling (HMM) is then used to identify the sequence of Doppler feature vectors as indicating a member of a particular class of targets. HMM is used to identify the sequence of Doppler feature vectors by assigning to each feature vector an occurrence probability by selecting a probability distribution or state from a set thereof associated with a class of targets, multiplying the occurrence probabilities together to obtain an overall probability, repeating for other probability distributions in the set to determine a combination of probability distributions giving highest overall probability for that class of target, then repeating for at least one other class of targets and selecting the target class as being that which yields the highest overall occurrence probability.