Abstract: A method of digitally evaluating the frequency and the phase of signals in the form of digitized samples, the method being wherein it comprises in succession:a stage during which numbers corresponding to the samples of the signal to be analyzed are processed in order to convert them into the form of an analytic signal whose real portion coincides with said signal to be analyzed; and, in parallel therewith:a stage of estimating the parameters to be analyzed which is performed in an overall manner on the basis of estimators and of selection criteria by working on the phase of the signal without using any operator of the Fourier transform type or any hypothesis test either separately or simultaneously; anda stage of estimating the differences between the real signal as taken in this way and the signal obtained from the estimated parameters, thereby making it possible to deliver data in digital form relating to the quality of the analyzed signal and to the reliability of the estimated values.

Abstract: A method of encoding and decoding image and/or audio signals includes an encoding step comprising previously preparing a quantization table which has pairs of a sampling interval and a quantized difference value, calculating a difference value between a given initial value and a sampling point succeeding to the initial value, comparing the calculated difference value with the quantization table, and searching the most matching quantization level. The search is started from the sampling point immediately succeeding to the initial value and then stopped upon finding of the matching quantization level. The index of that matching quantization level is transmitted. The decoding step comprises preparing the same quanitization table as used in the encoding step, determining the sampling interval and the quantized difference value based on the index received, and adding the quantized difference value to the initial value to obtain a decoded value.

Abstract: A method and implementing apparatus for low-bit rate speech band signal coding. An input signal in the speech band is represented by a pulse excitation sequence and a spectral parameter sequence over a frame of predetermined frame length using a selected one of a plurality of pulse determining processing modes. The selected pulse determining processing mode sequentially determines the amplitudes g.sub.i and locations m.sub.i of the pulses of the pulse excitation sequence on the basis of the amplitudes and locations of pulses in a previous frame. The selection process of determining which of the pulse determined processing modes to be used involves analyzing the input signal to produce a judgment signal d signifying the input signal as a voiced or an unvoiced signal, and selecting the pulse determining processing mode in response to the judgment signal d. The pulse excitation sequence and spectral parameter sequence are coded for transmission to a suitable receiver.

Abstract: In a method of and an arrangment for determining the start-point and end-point of a word signal in a speech signal consisting of isolated utterances, three adjacent windows are determined at each new digital value for the last arrived stored digital values, in which the central window contains the actual word signal. The length of this central window is varied for each digital value between a minimum and a maximum value, and a threshold value is determined from the two adjacent windows and is subtracted from the energy contained in the central window. Thus, the method and the apparatus always takes the overall speech signal into account instead of individual isolated portions so that a reliable end-point determination then is possible.

Abstract: A pattern recognition apparatus is constituted by a feature data extraction section, a RAM for storing feature data, a memory for storing standard pattern data, an arithmetic section for performing a similarity measurement calculation between input pattern data and the standard pattern data, based on a multiple similarity method, and a controller for recognizing an input signal, based on the calculation result from the arithmetic section.

Abstract: To improve the recognition of incoming speech signals in noise, the prestored templates of noise-free speech are modified to have the estimated spectral values of noise and the same signal-to-noise ratio as the incoming signal.

Abstract: For faster realtime analysis related to Discrete Fourier Transform (DFT), prestored pre-multiplied values of amplitude times cosine and sine functions are used. Method and apparatus for determining the content of a predetermined frequency bin f in the frequency spectrum of a signal and, in particular, a signal which occurs in a telephone network, includes the step of storing on a storage means:C.sub.k (x.sub.k,n)=x.sub.k *cos(2.pi.kn/N) and S.sub.k (x.sub.k,n)=x.sub.k *sin(2.pi.kn/N)for: k=0, . . . , (N-1); n=0, . . . , (n-1); and x.sub.k ranging over all the values of the set of M predetermined digital levels into which the amplitude of the signal is mapped, if the signal is a digital signal, or into which a sample of the amplitude of the signal is mapped, if the signal is an analog signal. N is a predetermined number of amplitudes or samples of the amplitude of the signal which are used in analyzing the frequency spectrum and n is a frequency index.

Abstract: In a multi-pulse excitation encoder, the number of pulses is controlled in accordance with an updated correlation signal compared to a predetermined value.

Abstract: A multi-user speech recognition system includes several speech recognizers which are shared between many user terminals. When a user terminal is activated, one of the recognizers is assigned to deal with voiced instructions received via that user terminal. This recognizer requests the respective "templates" from a memory device. A controller handles requests from the recognizers on a first-come first-served basis, and prevents later requests from interrupting earlier ones.

Abstract: An audio signal is initially represented by a series of high-resolution pulse code modulated (PCM) data. A lower rate series of representative values are extracted from the initial series of PCM data. Half of the lower rate is at an intermediate audio frequency so that the lower rate series encodes low frequency components of the audio signal. The PCM data are adjusted by offsetting in accordance with corresponding representative values and are then converted to a floating-point representation by extracting scale factor or exponents. The combination of the series of representative values and the floating-point data provides a rate-compressed representation of the audio signal which is capable of being decoded after transmission or storage to reproduce the audio signal without substantial noise, distortion or loss of dynamic range.

Abstract: A speech recognition apparatus is adapted to previously prepare a noise pattern in response to environmental noise prior to inputting a speech signal, evaluate a speech feature vector B.sub.i yielded by subtracting the noise pattern from a feature vector A.sub.i of the input speech upon inputting the speech signal thereafter, spectrum-normalize the speech feature vector B.sub.i, evaluate a local peaks vector by making use of binary-coding processing wherein only a component of a channel being the maximum of the spectrum-normalized vector in a direction of frequency is assumed to be "1", evaluate pattern similarity between an input pattern comprising a local peaks vector from the start point to the end point of the input speech and previously prepared reference patterns of a plurality of categories of the same format as the input pattern, and judge a category of a reference pattern being the maximum among the pattern similarities as a recognized result.

Abstract: A voice signal encoding and decoding apparatus and method which attenuates rapidly a voice signal when there occurs an error in transmission by giving increased robustness to the transmission error. It comprises an encoding section which includes a predictive filter for computing a predicted value of an input signal to the encoding section, and an encoder for encoding a difference signal obtained by subtracting from the input signal the predicted value predicted by the predictive filter; and a decoding section which includes a decoder for decoding a received difference signal, a detector which detects an error in the received signal, and a predictive synthesizing filter for computing the predicted value of the output signal from the sum of the signal decoded by the decoder and the computed prediction value of the output signal, and having a resonant frequency band width expanding when an error is detected by the error detector.

Abstract: In a speech encoder a Fourier transform of the speech is provided. The Fourier transform is equalized by normalizing the spectrum coefficients to a curve which approximates the shape of the spectrum. Both the curve and the equalized spectrum are encoded. Preferably, only a baseband of the normalized spectrum is encoded and that baseband is repeated in the decoder. The spectrum is normalized by scaling different regions (subbands) of the spectrum differently to flatten the spectrum.

Abstract: In a speech processor such as a speech recognizer, the problem of distortion of extracted features caused by adaptation of the input automatic gain control (AGC) during feature extraction is solved by storing the AGC's gain coefficient along with the energy level of each extracted feature. At the end of the sampling period the stored gain coefficients are set equal to the minimum stored coefficient and the associated energy levels adjusted accordingly. The AGC circuit may comprise a digitally switched attenuator under the control of a microprocessor performing the speech recognition.

Abstract: A method and apparatus for finding, by dynamic programing, combined patterns of standard patterns which are most similar to an input pattern by conducting a comparison collation between the input pattern expressed by a feature vector row and various combined patterns of memorized standard patterns expressed by feature vector rows corresponding to respective units of the words and the like to be recognized. In a dynamic programming calculation apparatus a comparison collation distance between a local pattern of the input pattern and an individual standard pattern is obtained, as a first step, by a method which is independent of a local length of the input pattern.

Abstract: A low cost speech recognition system generates frames of received speech having binary feature components. The received speech frames are compared with reference templates, and error values representing the difference between the received speech and the reference templates are generated. At the end of an utterance, if one template resulted in a sufficiently small error value, the word represented by that template is selected as the recognized word.

Abstract: Recognition of sound units is improved by comparing frame-pair feature vectors which helps compensate for context variations in the pronunciation of sound units. A plurality of reference frames are stored of reference feature vectors representing reference words. A linear predictive coder (10) generates a plurality of spectral feature vectors for each frame of the speech signals. A filter bank system (12) transforms the spectral feature vectors to filter bank representations. A principal feature vector transformer (14) transforms the filter bank representations to an identity matrix of transformed input feature vectors. A concatenate frame system (16) concatenates the input feature vectors of adjacent frames to form the feature vector of a frame-pair. A transformer (18) and a comparator (20) compute the likelihood that each input feature vector for a frame-pair was produced by each reference frame. This computation is performed individually and independently for each reference frame-pairs.

Abstract: This connected-speech recognition system uses a two-level hierarchical system, in which the higher-level (master) processor and one or more lower-level units (slaves) process, respectively, the most probably word sequence within a permitted grammar network, and the likelihood of individual words with the grammar network. The lower-level processing performs dynamic programming involving vector and matrix calculation and comparison, and processing speed is improved by an integrated processing unit which has simultaneous access to the external data memory as well as to a high-speed internal microinstruction ROM. One of the aforementioned units can also provide for performing an additional internal test function. The structure features two internal data buses and internal memories for more commonly used data and addresses, for enabling high-speed microinstruction performance and external memory access.

Abstract: A method of reconstructing, at the receiving end, digital data defining an encoded voice signal segment lost in transmission between the transmitter and the receiver in a transmission system wherein a low-frequency signal such as a residual baseband signal is derived at the transmitting end from the signal to be encoded and is then distributed among several sub-bands whose sampled contents are quantized separately. The reconstruction method includes a step of analyzing the received signal to detect any missing segment thereof and, as the case may be, to initiate an analysis, within each sub-band, of the segment(s) adjacent to the lost segment, so as to generate a term relating to the period T.sub.(k) of the signal so analyzed and to reconstruct a segment of signal of period T.sub.(k) intended to be substituted for the lost segment. Cyclic redundancy is used to detect errors, and modulo sequence numbering identification is used to detect loss of packet.

Abstract: An interrupt system provides interrupt signals to devices to be interrupted by indicating the presence of interrupts in a random access memory associated with each of the devices to be interrupted. The address of the interrupt signal that is written is assigned to a respective one of a plurality of addresses, each of which is assigned to a respective one of a plurality of interrupting devices and is indicative of the priority of the interrupt. The controller associated with each of the devices to be interrupted causes a scan of the associated memory and when an interrupt is detected, the address of the interrupt is sent to the interrupted device. The interrupted device then recognizes the interrupt by reason of its address and performs the appropriate interrupt routine. When an interrupt is written into the memory, a comparison is made of the address of the newly written interrupt with the address of the last scanned position.