Abstract: In a speech transmission or storage system using LPC parameters and a Baseband Signal derived from the prediction error signal, the synthesis excitation signal is formed from the baseband plus high-frequency regeneration, which is then spectrum-flattened for proper synthesis.

Abstract: In a speech analysis and synthesis (composing) method, a phoneme transition waveform is generated to smoothly couple phoneme PHA to PHB having different pitch periods. The method involves obtaining the differential value (difference phoneme .DELTA.PH or difference phoneme .DELTA.PHI and difference clock or pitch period), and in composition merely adding phoneme PHA and difference phoneme .DELTA.PHI and clock period and difference clock period.

Abstract: A speech recognition system for an automotive vehicle derive a spoken instruction start signal and a spoken instruction end signal when a smoothed spoken instruction signal exceeds or drops below a predetermined threshold level representing the intensity of the background noise for more than first and second predetermined time periods, respectively. Noise is determined by converting the output of a microphone transducing the spoken instruction into a single polarity signal that is smoothed with a long time constant. The single polarity variation is also smoothed with a shorter time constant. The signals with the long and short constants are applied to a comparator that derives a bi-level output signal. In response to transitions in first and second directions of the bi-level signal lasting for first and second durations, the start and end signals are respectively derived.

Abstract: Speech and melody information are separately inputted and stored. The speech timing is modified (corrected) to alignment with the melody.

Abstract: In a speech recognition system, similarity calculations between speech feature patterns are reduced by stopping similarity calculations for any one reference pattern when a frame in the pattern fails to exceed a corresponding similarity threshold.

Abstract: A speech recognition system for an automotive vehicle sequentially activates at least one device having plural different operating modes in accordance with a plurality of spoken instructions without repeating the same phrase. To adjust fender mirrors, for instance, first the driver says "Mirror" while depressing a recognition switch to stop the moving mirror; secondly the driver says "Right, horizontally", thirdly the driver says "Right, vertically" after having depressed a reset switch; lastly the driver says "Stop". The speech recognition system comprises a reset switch, at least one device instruction phrase memory, plural operation-mode instruction phrase memories, and a reference pattern memory selector for selecting the device memory or an appropriate operation-mode memory, in addition to or in place of a section of a conventional speech recognizer.

Abstract: Phoneme recognition uses the silence-phoneme and phoneme-phoneme transition spectral information rather than the phoneme information itself. The transition detector features first and second differences in level for each frequency band.

Abstract: A continuous speech recognition system determines the similarity between input patterns and reference patterns over time such that similarities between previously spoken speech patterns and reference patterns are determined while speech continues to be spoken. Degrees of dissimilarity at arbitrary reference pattern word times are determined asymptotically and are recorded. The minimum degree of dissimilarity is determined and the corresponding word is categorized. Recognition decisions are ultimately made in reverse chronological order.

Abstract: A voice-controlled elevator security system. Access to an elevator system servicing a plurality of landings of a structure is controlled by comparing a voice signal of a potential user of the elevator system with stored voice signals of all authorized users. The comparison is accomplished by first digitizing the potential user's voice signal and comparing it, on a bit-by-bit basis, with a digital representation of the voice signals of all authorized users. If the comparison satisfies an established criterion, the potential user is determined to be an authorized user. Once this determination has been made, the authorized user can initiate car call signals for those landings to which the authorized user is permitted access. An authorized user can also temporarily modify the previous digital representation of the voice signal when the authorized user is suffering from laryngeal indisposition.

Abstract: In this speech recognition system, a set of templates for each phoneme includes clusters of speech patterns based on two speech features: "physical" features (formant spectra of men versus women) and "utterance" features (unvoiced vowels and nasalization), derived from a plurality of reference speakers.

Abstract: In a TASI system using dynamic subband allocation and BCPCM (Block Companded PCM), losses due to freeze-out or delay are avoided by use of a multi-rate macro-frame format, wherein switching to a lower rate (24, 16, or 8 Kbs) for one or more of the input channels is accommodated by simply dropping one or more sections of each signal frame, since the lower rate bits are distributed over the entire frame.

Abstract: Speech signal is decided voiced or unvoiced by a sequence of unilateral decisions: a first test decides "unvoiced" if standardized energy E.sub.s is below a threshold, or "ambiguous" if above the threshold whereby a second test decides "unvoiced" if the number of zero crossings ZC is above a threshold, and ambiguous if below the threshold. Up to six criteria may be so tested as ambiguous before a "voiced" decision is made.

Abstract: A method and apparatus for use in combination with a communications system, such as, a two-way radio which will simulate human voice at the level of a common microphone has been devised wherein the host radio has a transmit and receive mode for the transmission and reception of voice messages. The system has a memory device with at least one pre-encoded message permanently stored therein and a voice generator operative to convert the pre-encoded message into electrical analog voice signals which simulate the human voice. The device is programmable to extract one message from the memory and repetitively transmit the message at spaced time intervals to the voice generator. A relay control interconnects the voice generator and host radio for broadcasting the voice signals from the voice generator and is also capable of automatically converting the host radio to the receive mode for the reception of messages from an external source during the spaced time intervals between the transmission of the voice signals.

Abstract: A device for providing real time speech compression and expansion and method therefor which enables speech data, sampled at a predetermined rate, and consisting of a preselected number of bits per sample, to be compressed in real time by a factor of approximately 2:1, and upon decomposition, to be essentially indistinguishable from the original digitized speech. In particular, the apparatus employs a method which involves the digitizing of the input data, determining the maximum amplitude of the data over a short given period, compressing the data by dividing the largest amplitude constant from the data over the given period to obtain a data sample, and storing the reduced data sample and corresponding amplitude constant or gain factor in a suitable storage medium. In the preferred embodiment, the storage medium is a rotating memory or a recirculating shift register.

Abstract: In a frequency conversion apparatus, the input signal is sampled and stored at one clock frequency, then read out at another clock rate and passed through a filter whose attenuation characteristics are a function of the second clock rate.

Abstract: A formant-based multi-path speech synthesizer is reconfigurable to use common elements for different types of sounds. A pitch generator or a noise generator is connected to a glottal signal path for vowel or aspirate sound, respectively. The pole and zero filters normally part of the fricative signal path which produces fricative or stop sounds are used in the glottal signal path for nasal sounds. The output of a spectral filter in the glottal signal path bypasses the cascaded formant filters and is connected directly to the glottal path attenuator for voice bar sounds. The output of the first formant filter is rectified and modulates the noise signal in the fricative path and the glottal and fricative signal paths are summed for voiced fricative sounds. To minimize spectral tilt, the third and fourth formant filters and the pole filter are peak filters. The glottal path attenuator is at the end of the glottal signal path to maximize the signal-to-noise ratio.

Abstract: Spoken words are recognized in a sequence of steps: first finding candidate words in the form of mono-syllabic sequences; then associating (correcting) each mono-syllabic sequence to more speech-like sounds, then comparing with pre-stored dictionary words for final recognition.

Abstract: Speech synthesizer and a computer system having the speech synthesizer operably coupled thereto to provide speech capability for the computer system. The speech synthesizer is capable of electronically synthesizing human speech from coded speech data including parameters as stored either in a solid state memory on a permanent basis or alternatively as temporarily stored in another memory, wherein the coded speech data is made available from an external source, such as a central processing unit of a commercial or home-type computer, as coupled to the speech synthesizer.

Abstract: In a communication system an input predictive type signal is analyzed in successive time intervals to generate a set of prediction signals for the interval. A predictive residual signal is produced jointly responsive to the input signal and the interval prediction signals. The predictive residual signal is quantized and encoded. The quantization includes dividing the predictive residual signal into a plurality of distinct portions and selectively quantizing the distinct portions to improve intelligibility.

Abstract: An arrangement for endpoint detection improves speech recognition accuracy and lowers rejection rates by developing an ordered list of endpoint candidates. A triple thresholding technique defines energy signal pulses. The energy pulses are combined according to predetermined criteria to form the endpoint candidates.