Abstract: A method for analyzing a speech signal to isolate speech and nonspeech portions of the speech signal is provided. The method is applied to an input speech signal to determine boundary values locating isolated words or groups of words within the speech signal. First, a comparison signal is generated which is biased to emphasize components of the signal having preselected frequencies. Next, the system compares the comparison signal with a threshold level to determine estimated boundary values demonstrating the beginning and ending points of the words. Once the estimated boundary values are calculated, the system adjusts the boundary values to achieve final boundary values. The specific amount of adjustment varies, depending upon the amount of noise present in the signal. The final pair of boundary values provide a reliable indication of the location and duration of the isolated word or group of words within the speech signal.

Abstract: A system including a touch pressure-sensitive transducer and a computer responsive thereto for producing a sentic cycle and for recording touch expression in the course of which cycle different emotions are expressed and generated by applying appropriate finger pressure to the transducer actuator. Stored in the memory of the computer is a set of words representing the different emotions, the computer being programmed to sequentially select these words at timed intervals and to audibly reproduce the selected word. Each word is followed by a series of time-spaced audible start clicks, each commanding the subject when to express with finger pressure on the transducer actuator. The signals yielded by the transducer reflecting vector components of the applied finger pressure are processed in the computer whose display terminal then presents on its screen a sentogram, the shape of which characterizes the emotion sensed by the transducer.

Abstract: A method of screening communication functions in a human subject comprises (a) presenting a verbal auditory stimulus to the subject, and then (b) scoring a response to the verbal auditory stimulus, with the response being an expressive response, a receptive response, or both. These steps are then cyclically repeated to provide an evaluation of the subject's response to a plurality of verbal auditory stimuli. Once the evaluation is complete, the evaluation is used to determine whether the subject should receive further diagnostic testing. In a preferred embodiment of the invention, subjects are deliberately confounded during the receptive portion of the test.

Abstract: Coding digitized audio signals includes dividing an audio signal, consisting of a continuous sequence of sample values, into successive blocks of equal length, and performing overlapping windowing. The blocks are transformed into complex Fourier coefficients by means of a discrete Fourier transform, which complex Fourier coefficients are decomposed into magnitude values and phase values. The phase values are quantized with a linear quantization characteristic which becomes coarser going from low toward high frequencies. The magnitude values are combined into frequency bands which are oriented with regard to predetermined critical bands and become wider toward high frequencies.

Abstract: A message synthesizer circuit includes a compressed message data memory as a message source for storing a plurality of compressed message data of message, each corresponding to a message code specifying a message to be emitted as a synthesized message. An input message code selector converts a message code signal into the count of a ring counter by taking, as its inputs, a count output emitted from the ring counter, with the total number of these compressed message data corresponding to its maximum count number, a message code signal for specifying a message to be emitted and an input message code selector signal for setting a randomizing condition for randomly altering the message code signal. The system controller reads out the compressed message data corresponding to the random message code from the compressed message data memory, which is then converted into a specific synthesized message.

Abstract: Analog speech signals are coded as digital signals before transmission over a transmission medium and then are decoded at their destination. The coder is of the linear predictive type (LPC) and includes an LPC analyzer for adjusting an analysis filter, which receives the digital signal and generates a residual signal representative of error content. The parameters by which the filter is adjusted by the analyzer and the residual signal together represent the digital signal. The residual signal is split into segments and, per segment, several first pulse train signals are generated, each having a different starting time position within the segment. The first pulse train signal which is most closely related to the residual signal is selected and compared to second pulse train signals stored in a codebook.

Abstract: The invention relates in general to digital encoding and decoding of information. More particularly, the invention relates to efficient implementation of digital analysis and synthesis filter banks used in encoding and decoding. The invention permits the length of a digital transform used to implement critically-sampled analysis and synthesis filter banks to be adaptively selected.

Abstract: A noise signal prediction system includes a signal detector for receiving a mixed signal having a voice signal and a background noise signal and for detecting the presence and absence of the voice signal contained in the mixed signal. A noise level detector is provided for detecting an actual noise level at each sampling cycle during the absence of the voice signal. A storing circuit stores the noise levels for a predetermined number of past sampling cycles. A predicting circuit predicts a noise level of a next sampling cycle based on the stored noise levels in the storing circuit. The storing circuit receiving and stores the actual noise levels during the absence of the voice signal, but stores the predicted noise levels during the presence of the voice signal.

Abstract: A speech detection apparatus capable of reliably detecting speech segments in audio signals regardless of the levels of input audio signals and background noises. In the apparatus, a parameter of input audio signals is calculated frame by frame, and then compared with a threshold in order to judge each input frame as one of a speech segment and a noise segment, while the parameters of the input frames judged as the noise segments are stored in the buffer and the threshold is updated according to the parameters stored in the buffer. The apparatus may utilize a transformed parameter obtained from the parameter, in which the difference between speech and noise is emphasized, and noise standard patterns are constructed from the parameters of the input frames pre-estimated as noise segments.

Abstract: A method and apparatus for modeling words based on match scores representing (a) the closeness of a match between probabilistic word models and the acoustic features of at least two utterances, and (b) the closeness of a match between word models and the spelling of the word. A match score is calculated for a selection set of one or more probabilistic word models. A match score is also calculated for an expansion set comprising the probabilistic word models in the selection set and one probabilistic word model from a candidate set. If the expansion set match score improves the selection set match score by a selected nonzero threshold value, the word is modelled with the word models in the expansion set. If the expansion set match score does not improve the selection set match score by the selected nonzero threshold value, the word is modelled with the words in the selection set.

Abstract: A voice log-in system is based on a person's spoken name input only, using speaker-dependent acoustic name recognition models in a performing speaker-independent name recognition. In an enrollment phase, a dual pass endpointing procedure defines both the person's full name (broad endpoints), and the component names separated by pauses (precise endpoints). An HMM (Hidden Markov Model) recognition model generator generates a corresponding HMM name recognition model modified by the insertion of additional skip transitions for the pauses between component names. In a recognition/update phase, a spoken-name speech signal is input to an HMM name recognition engine which performs speaker-independent name recognition--the modified HMM name recognition model permits the name recognition operation to accommodate pauses between component names of variable duration.

Abstract: Disclosed is an Hidden Markov Model (HMM) training apparatus in which a capacity for discriminating between models is taken into consideration so as to allow a high level of recognition accuracy to be obtained. A probability of a vector sequence appearing from HMMs is computed with respect to an input vector and continuous mixture density HMMs. Through this computation, the nearest different-category HMM, with which the maximum probability is obtained and which belongs to a category different from that of a training vector sequence of a known category, is selected. The respective central vectors of continuous densities constituting the output probability densities of the same-category HMM belonging to the same category as that of the training vector sequence and the nearest different-category HMM are moved on the basis of the vector sequence.

Abstract: On carrying out connected word recognition in compliance with a regular grammar and in synchronism with successive specification of input feature vectors of an input pattern, one of an n-th word (n) of first through third occurrence (n(1) to n(3)) is selected as the n-th word of selected occurrence in an i-th period in which an n-th input feature vector is specified. The n-th word appears as the first through the third occurrence in transition rules specified by the grammar. In the i-th period, distances are calculated, only for the n-th word of the selected occurrence, between the input feature vector assigned to the i-th period and reference feature vectors for the n-th word.

Abstract: A method and apparatus for utilizing the sound and pattern recognition capabilities of the human auditory system, which takes information contained in typical instrumentation signals in the form of amplitude, frequency, and time characteristics, and converts this information into sound qualities and characteristics which are recognizable by the human listener. The method and apparatus digitizes the analog amplitude, frequency, and time information, and selects appropriate sound characteristics into which it may encode this information in standardized form that is recognizable to the human listener. The method allows for the testing or inspection of materials using nondestructive evaluation techniques in a manner that allows the tester to interpret the information provided by the testing system, either exclusively through his auditory senses, or through his auditory senses in conjunction with visual indicators.

Abstract: A coding apparatus is disclosed, in which coding can be achieved by simply storing, in a memory, the results of necessary coding operations including the feedback of the quantization error and reading out the data from an address of the memory specified by input data. Consequently, an encoded code can be obtained at a far higher speed than in a case where operations are performed by individual operation circuits for coding and decoding.

Abstract: A machine for neural computation of acoustical patterns for use in real-time speech recognition, comprising a plurality of analog electronic neurons connected for the analysis and recognition of acoustical patterns, including speech. Input to the neural net is provided from a set of bandpass filters which separate the input acoustical patterns into frequency ranges. The neural net itself is organized into two parts, the first for performing the real-time decomposition of the input patterns into their primitives of energy, space (frequency) and time relations, and the second for decoding the resulting set of primitives into known phonemes and diphones. During operation, the outputs of the individual bandpass filters are rectified and fed to sets of neurons in an opponent center-surround organization of synaptic connections ("on center" and "off center"). These units compute maxima and minima of energy at different frequencies.

Abstract: In speech recognition and speech coding, the values of at least two features of an utterance are measured during a series of time intervals to produce a series of feature vector signals. A plurality of single-dimension prototype vector signals having only one parameter value are stored. At least two single-dimension prototype vector signals having parameter values representing first feature values, and at least two other single-dimension prototype vector signals have parameter values representing second feature values. A plurality of compound-dimension prototype vector signals have unique identification values and comprise one first-dimension and one second-dimension prototype vector signal. At least two compound-dimension prototype vector signals comprise the same first-dimension prototype vector signal. The feature values of each feature vector signal are compared to the parameter values of the compound-dimension prototype vector signals to obtain prototype match scores.

Abstract: In a continuous speech recognizer which includes at least, one acoustic expert and one linguistic expert which generate respective scores, a method is disclosed for adjusting the relative weighting to be applied to those scores employing training data utilizing the words to be recognized in multiple word phrases. Multiple word test phrases are applied to the acoustic expert to determine, for each phrase, plural multi-word hypotheses each having corresponding cumulative scores. The linguistic expert generates corresponding cumulative linguistic scores. An objective function is calculated for each test phrase having a value which is variable as a function of the difference between the combined score of any correct hypothesis and that of the most easily confused incorrect hypothesis. The objective function values are cumulated and a gradient descent procedure is used to adjust the relative weighting of the acoustic and linguistic scores in obtaining a combined score.

Abstract: A speech coding apparatus and method for use in a speech recognition apparatus and method. 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. A plurality of prototype vector signals, each having at least one parameter value and a unique identification value are stored. The closeness of the feature vector signal is compared to the parameter values of the prototype vector signals to obtain prototype match scores for the feature value signal and each prototype vector signal. The identification value of the prototype vector signal having the best prototype match score is output as a coded representation signal of the feature vector signal. Speaker-dependent prototype vector signals are generated from both synthesized training vector signals and measured training vector signals.

Abstract: A voice animation system decomposes pre-recorded samples of actual speech into basic segments to derive speech patterns of a particular speaker to provide parameters and coefficients for use in a text-to-speech synthesizer to artificially synthesize human quality speech with unlimited vocabulary in the voice of the person who provided the pre-recorded samples. The pre-recorded speech samples are further processed to add desired inflection and other auditory effects to create high-quality animated or artificial voices.