Abstract: Speech pattern data representing speech of a plurality of speakers are stored in a pattern storage section in advance. Averaged pattern data obtained by averaging a plurality of speech pattern data of the first of the plurality of speakers are obtained. Data obtained by blurring and differentiating the averaged pattern data are stored in an orthogonalized dictionary as basic orthogonalized dictionary data of first and second axes, respectively. Blurred data and differentiated data obtained with respect to the second and subsequent of the plurality of speakers are selectively stored in the orthogonalized dictionary as additional dictionary data having new axes. Speech of the plurality of speakers is recognized by computing a similarity between the orthogonalized dictionary formed in this manner and input speech.

Abstract: Method and apparatus for identifying telephone signals which are produced by a human voice. The method includes determining a measure of the energy of predetermined frequency bins of the frequency spectrum of the signal for a frame comprised of a predetermined number of consecutive samples of the signal, if the signal is an analog signal; averaging the measure of the energy of the predetermined frequency bins over a predetermined number of frames to provide a frame group average energy measure; determining the sum of the absolute value of the difference between the frame group average energy measure for all pairs of frame groups which are consecutive in time; and comparing the sum with a predetermined amount to identify the signal as having been produced by a human voice if the sum exceeds the predetermined amount.

Abstract: An efficient pruning method reduces central processing unit (CPU) loading during real time speech recognition by instructing the CPU to compare a current state's previously calculated probability score against a predetermined threshold value and to discard hypothesis containing states with probability scores below such threshold. After determining that the current state should be kept, the CPU is directed to locate an available slot in the scoring buffer where information about the current state is then stored. The CPU locates an available slot by comparing the current time-index with the time-index associated with each scoring buffer slot. When they are equal, the slot is considered not available; when the current time-index is greater, the slot is considered available. After the information about the current state is stored, the CPU then sets the current state's backpointer to point at the start state of the current best path if the current states represents a completed model.

Abstract: An apparatus including a measuring head, equipped with light sources for lighting one part of a bodywork of which the color is to be reproduced with a receiver for picking up the light reflected in each of a plurality of preset wavelength in order to produce a signal representing the reflectance of the surface in each wavelength. The apparatus also includes a device for calculating, for each wavelength, the reflectance for any combination of basic shades or bases from among a predetermined set off bases, and apparatus for continuously comparing the measured reflectance with the calculated reflectance in order to find a suitable paint formulation. A starting formulation given by the particular combination of bases corresponding to the color reference of the bodywork to be repaired may be modified, until the difference between the calculated reflectance values and measured reflectance values reaches a minimum or becomes smaller than a preset threshold.

Abstract: In a neural network, input neuron units of an input layer are grouped into first through J-th input layer frames, where J represents a predetermined natural number. Intermediate neuron units of an intermediate layer are grouped into first through J-th intermediate layer frames. An output layer comprises an output neuron unit. Each intermediate neuron unit of a j-th intermediate layer frame is connected to the input neuron units of j'-th input layer frames, where j is variable between 1 and j and j' represents at least two consecutive integers, one of which is equal to j and at least one other of which is less than j. Each output neuron unit is connected to the intermediate neuron units of the intermediate layer. For recognition of an input pattern represented by a time sequence of feature vectors, each consisting of K vector components, where K represents a predetermined positive integer, each input layer frame consists of K input neuron units.

Abstract: The apparatus includes circuitry for obtaining a video image of an individual's face, circuitry for electronically locating and tracking a first feature, such as the nostrils, of the facial image for use as reference coordinates and circuitry responsive to the reference coordinates for locating and tracking a second facial feature, such as the mouth, of the facial image with respect to the first feature. By tracking the location of the nostrils, the apparatus can follow the movement of the mouth, and thereby automatically recognize speech. In a preferred embodiment, the video image is grayscale encoded and the raster lines are smoothed to eliminate noise. The transitions between gray levels of the smoothed image are encoded and the resulting transition code is used to form a contour map of the image from which region parameters are computed which can be compared against stored speech templates to recognize speech.

Abstract: The system utilizes a linear predictive coding (LPC) analyzer, an Attenuator, a line spectrum pair (LSP) analyzer, a reference pattern memory and a pattern matching device. The LPC analyzer derives LPC parameters from an input speech signal. The LPC parameters are attenuated in the attenuator and fed to the LSP analyzer for deriving LSP parameters which are in turn fed to the pattern matching device. The reference pattern memory stores a plurality of reference patterns composed of a sequence of LSP parameters for a variety of predetermined speech samples. The pattern matching device is connected to the LSP analyzer and the reference pattern memory to select the reference pattern which most closely resembles the input pattern from the LSP analyzer and to provide a label code as an output thereof. On the decoding side, a decoder is responsive to the label for generating LPC parameters corresponding to the reference pattern of the label.

Abstract: A character voice communication system including high efficiency voice coding system for encoding and transmitting speech information at a high efficiency and a voice character input/output system for converting speech information into character information or receiving character information and transmitting speech or character information are organically integrated. A speech analyzer and a speech synthesizer are shared by both the voice coding and the voice character input/output systems. Communication apparatus is also provided which allows mutual conversion between speech signals and character codes.

Abstract: Comparatively macroscopic phonemic units consisting of continuous sounds are used as standard patterns in continuous dynamic programming. In matching intervals between these standard patterns and input patterns, a sampled pattern is labeled according to the phonemes and phonemic particles which form the standard patterns. In the matching intervals, the labeled pattern is called a matching history. The matching history of a standard pattern when the continuous dynamic programming value is below a certain threshold value is used to recognize the phonemes of the input patterns.

Abstract: A speech processing system having an encoder comprising apparatus for receiving successive samples of PCM (pulse code modulated) encoded speech signals and apparatus, for applying sequential groups of the PCM encoded speech signals as primary vector signals to an encoder code book memory for selecting code words stored in the memory most closely approximating the vector signals. Apparatus is included for outputting to an output line the selected code words at a first bit rate. Further apparatus connects the selected code words to converted vector signals. The primary vector signals and converted vector signals are compared and difference signals result. The difference signals are quantized error signals are provided thereby.

Abstract: In the transmission of audio signals, the audio signal is digitally represented by use of quadrature mirror filtering in the form a plurality of spectral sub-band signals. The quantizing of the sample values in the sub-bands, e.g. 24 sub-bands, is controlled to the extent that the quantizing noise levels of the individual sub-band signals are at approximately the same level difference from the masking threshold of the human auditory system resulting from the individual sub-band signals. The differences of the quantizing noise levels of the sub-band signals with respect to the resulting masking threshold are set by the difference between the total information flow required for coding and the total information flow available for coding. The available total information flow is set and may then fluctuate as a function of the signal.

Abstract: A real-time vector adaptive predictive coder which approximates each vector of K speech samples by using each of M fixed vectors in a first codebook to excite a time-varying synthesis filter and picking the vector that minimizes distortion. Predictive analysis for each frame determines parameters used for computing from vectors in the first codebook zero-state response vectors that are stored at the same address (index) in a second codebook. Encoding of input speech vectors s.sub.n is then carried out using the second codebook. When the vector that minimizes distortion is found, its index is transmitted to a decoder which has a codebook identical to the first codebook of the decoder. There the index is used to read out a vector that is used to synthesize an output speech vector s.sub.n.

Abstract: Apparatus for analyzing distortion levels of samples which occur in data compression systems of the type that may be integrated into a data compression system or used as a stand alone analyzer. The apparatus includes a processor that is coupled to the data compression system to provide the distortion analyzing apparatus with necessary input bit allocation and power level data for each sample. Memory storage is coupled to the processor for supplying predetermined unique quantization error values corresponding to each transform coefficient representing samples of digital data in a frame of digital data to be compressed.

Abstract: A circuit arrangement for storing a speech signal in a digital speech memory (SP) and reproducing the stored signal comprises among other things a control circuit (SS). The control circuit produces control commands and clock signals for the individual modules of the arrangement. The speech signal to be stored is available in one of various predetermined codes. In order to enable storing speech signals of considerable lengths, it is provided to write the speech signal into the digital speech memory (SS) in a preferred digital code. For this purpose, input converters (EW1, EW2, EW3, EW4) are used which convert the code of the speech signal to be stored from the one it has into the preferred digital code, if the signal is not already available in this code. Multiplexing means (M1, M2), controlled by the control circuit (SS), further convey the converted speech signal to the input of the digital speech memory (SP).

Abstract: An apparatus for generating synthesized voice from a bar code includes a voice encoding unit of a voice data input unit, a composition data analyzer, and an analyzed information library arranged independent of a voice response unit of a control information generator and voice composer. Analyzed information is written into the information store means in the voice encoding unit.

Abstract: In a device for and a method of encoding a first signal (f.sub.0), for example a speech parameter such as the pitch, as a function of time, to form a second signal (FIG. 2), a third signal (k) is derived from the first signal. The third signal is a measure of the curvature of the first signal as a function of time. The extrema (such as k(t.sub.1) in FIG. 1b) in this third signal are determined and the second signal is generated in the form of a sequence of information blocks (B.sub.1, B.sub.2, . . . ), of which one information block (such as B.sub.3) contains time information corresponding to the instant (t.sub.3) at which an extremum occurs in the third signal.

Abstract: A voice operated switch employs digital signal processing techniques to examine audio signal frames having harmonic content to identify voiced phonemes and to determined whether the signal frame contains primarily speech or noise. The method and apparatus employ a multiple-stage, delayed-decision adaptive digital signal processing algorithm implemented through the use of commonly available electronic circuit components.

Abstract: A method and system for providing adaptive interactive command response to a user, in which the user may protest upon incorrect recognition by the system of a command given by the user in response to question from the system. If the number of such protests reaches a value which exceeds a predetermined level of at least two protests, the decision model used for command recognition by the system is adjusted so that during a certain time interval (T) thereafter a more explicit indication is given to the user concerning one or more of the commands which are recognized. In the normal case, when such threshold value is not exceeded, no indication is given to the user of commands which have been recognized and the system proceeds directly to the next question.

Abstract: A sub-band speech coding arrangement divides the speech spectrum into sub-bands and allocates bits to encode the time frame interval samples of each sub-band responsive to the speech energies of the sub-bands. The sub-band samples are quantized according to the sub-band energy bit allocation and the time frame quantized samples and speech energy signals are coded. A signal representative of the residual difference between the each time frame interval speech sample of the sub-band and the corresponding quantized speech sample of the sub-band is generated. The quality of the sub-band coded signal is improved by selecting the sub-bands with the largest residual differences, producing a vector signal from the sequence of residual difference signals of each selected sub-band, and matching the sub-band vector signal to one of a set of stored Gaussian codebook entries to generate a reduced bit code for the selected vector signal.

Abstract: An object sensor (or event detector) apparatus detects distinctive characteristics (such as color) of otherwise physically similar objects. Respective indices or counters are incremented in response to the various detections. The event detector (object sensor) may take various forms such as a conveyor belt production monitor, but is especially useful as an electronic scorekeeper for a competitive game which simultaneously keeps score for two or more different players by distinguishing between the different colors (or other characteristics) of balls or other objects each player uses to play a given game. A solitary mode also keeps score for a single player within a given time period. In this selectable alternative mode, the event detector apparatus records accumulated points and displays the time remaining in a given time period which defines a game time period.