Abstract: A process for deriving voice pitch related delay values M to tune a Long-Term Prediction (LTP) filter to be used in an LTP based speech coder converting a speech derived digital signal r(n) into a lower bit rate signal, said filter being provided with a variable length delay line y(n) fed with a reconstructed signal r'(n). The process includes splitting r(n) into segments and each segment into sub-segments; then cross-correlating the first current r(n) sub-segment with a previously reconstructed segment and sorting the cross-correlation values for peak location, whereby a first delay value M1 is derived and used to tune the filter. Then, said M1 is used to compute sample indexes n for a predefined number of samples located about M1/p, . . . , M1, 2M1, . . . , pM1 and repeating cross-correlation and sorting operations to derive M2 and so on up to a full segment length (e.g. 160 samples). Then the process is started all over again.

Abstract: A speech coding and decoding apparatus for use in linear predictive coding of a speech signal includes coding apparatus having a pitch analyzing element for separating an analysis frame into one or more blocks and calculating the strength of correlativity between pitch periods of a residual waveform in each block. A residual partially compresing element compresses the time axis of the residual waveform in a block having a high correlativity strength. A residual quantizing element quantizes the compressed residual waveform while preferentially allotting quantization allotment bits to the compressed portion of the residual waveform. Decoding apparatus includes a residual inverse quantizing element for inversely quantizing the residual waveform by the same bit allotment. A residual partially expanding element expands the compressed portion of the residual waveform to its original length and supplies it to a linear predictive synthesis filter.

Abstract: In an encoder device for encoding a sequence of digital speech signals classified into a voiced sound and an unvoiced sound into a sequence of output signals, by the use of a spectrum parameter and pitch parameters, at every frame having N samples where N represents an integer, a judging circuit judges whether the digital speech signals are classified into the voiced sound or the unvoiced sound to produce a judged signal representative of a result of judging. A processing unit processes the digital speech signals in accordance with the judged signal to selectively produce a first set of primary sound source signals and a secondary sound source signals. The first set of primary sound source signals are produced when the judged signal represents the voiced sound and are representative of locations and amplitudes of a first set of excitation multipulses calculated at every frame.

Abstract: A conventional speech recognition network finite-state automaton, which follows regular grammar rules, is improved by adding subnetworks tapped into the original network at call and return points, whereby context-free grammar rules may be used, with avoidance of infinite loop response of a recurrent expression.

Abstract: A class of methods and related technology for determining the phase of each harmonic from the fundamental frequency of voiced speech. Applications of this invention include, but are not limited to, speech coding, speech enhancement, and time scale modification of speech. Features of the invention include recreating phase signals from fundamental frequency and voiced/unvoiced information, and adding a random component to the recreated phase signal to improve the quality of the synthesized speech.

Abstract: A system for voice coding based on vector quantization has an apparatus in which a distribution area of parameters representative of a voice is divided into a plurality of domains so that one vector (code vector) may correspond to one domain, an apparatus for representing individual code vectors by codes specific thereto, an apparatus for converting an input voice into a vector and determining membership functions by numerically expressing the distance between the nearest code vector and each of the predetermined number of neighboring vectors, and an apparatus for transmitting, as fuzzy vector quantization information, a code of the nearest code vector and the membership functions.

Abstract: A low-overhead method of protecting multi-pulse speech coders from the effects of severe random or fading pattern bit errors combines a standard error correcting code (convolutional rate 1/2 coding and Viterbi trellis decoding) for protection in random errors with cyclic redundancy code (CRC) error detection for fading errors. Compensation for detected fading errors takes place within the speech coder. Protection is applied only to the perceptually significant bits in the transmitted frame.

Abstract: The detection of DTMF tones is improved in a three-step process: first testing even-numbered samples; then testing odd-numbered samples, and finally testing for tone quality (time interval and frequency stability).

Abstract: A method and apparatus are disclosed for recognizing spoken commands uttered by a user and for generating responsive control signals once the command is recognized. In accordance with this disclosure the audio signal is converted into a series of count bytes representing the time between the audio signal zero crossings, and all the count bytes of the full command are then segmented into equal temporal groups and sorted within each segment into a set of frequency class intervals which are based on a computation of substantially equal byte activity in all the words comprising the command lexicon. In this manner, lower and higher frequency groups are selected for equal significance. The uttered words are then compared against stored words similarly transformed according to segment and frequency interval and if the comparison conditions are satisfied the command is executed; if not, an indication is provided to the user to repeat the command. Segmenting produces a segment period.

Abstract: Accelerated throughput of voice messages is accomplished by first filtering a normal 50-3000 Hz voice signal to a reduced bandwidth 350-1650 Hz (with loss of identifying features), the reduced signal then accelerated X3 and translated downward to bandwidth 50-3950 Hz for transmission. The receiver recovers the 350-1650 Hz signal.

Abstract: A muting circuit in a digital audio system having a digital signal processor, a first latch, a second latch, a comparator for comparing data in the first and second latches, an address encoder, a counter, a memory, a divider, a multiplier and a switching circuit, wherein disturbing beat noises generated during the turning off of power to the system or generated null data pop noises generated in response to external influences or internal circuitry influences are muted.

Abstract: Power information from an input signal is used to detect the presence of speech. A reference point is established by the power information as the moment of detection of the start of speech. The end of a processing period for speech recognition is determined by a power information occurring after end of speech, so that feature parameters are extracted from the signal during the processing period having start and end points determined by the power information. Standard speech patterns of particular preset words are determined and similarities between the extracted feature parameters and the standard patterns are calculated and mutually compared. The selected preset word corresponds to a maximum of the similarities obtained during the processing period selected in accordance with the power information. The selected word is then outputted as the recognition result.

Abstract: A method and a device for producing an imitative animal's voice to embellish a music. The animals voice is analyzed and approximated into a waveform represented exclusively by HIGH/LOW, and the time data X of each group of consecutive intervals of the same state are stored in a first ROM. The data X are stored in the consecutive addresses of a first read only memory ROM. When the ROM receives a pulse from a first address counter, the datum X stored in the mth address of the ROM will be sent to a first divider means if the address count of the first address counter is m. To further melodize the imitative animal's voice, the clocks from a first clock generator are compressed or expanded. The data Y, Z of the notes of the desired melody are stored in the consecutive addresses of a second ROM to respectively control the average (or apparent) pitch of the produced imitative voice and the duration of the voice at a given pitch.

Abstract: In a speech coding method and system in which a speech signal is analyzed in each frame so as to be separated into spectral envelope information and excitation information and both of the information are coded, each frame is divided into a plurality of sub-frames and a pulse of the maximum-amplitude is extracted from pulses within each sub-frame in order to provide large-amplitude pulses from each frame, thereby greatly reducing the number of pulse extracting processing steps.

Abstract: Time-serial pattern data of feature parameters representing the frequency feature of input speech is obtained from an input speech signal and is output from a frequency analyzer. This time-serial pattern data is plotted on the frequency and time base axes in relation to a power level. A Central Processing Unit detects that the feature parameters continuously exceed a predetermined value for a predetermined period of time or more. The detection output is supplied to an amplifier, and its input gain is controlled to be an optimal value.

Abstract: An subband coding system encoder is provided for improving the quality of a reconstructed speech signal. Briefly, the improved subband coding system encoder includes means for improving the selection of the waveform/noise fill gain factor. The improved selection method, according to the invention, is based on the quasi-stationary characteristic of speech, that is, that the short-time frequency spectrum of a speech signal varies slowly with time. Moreover, the amount of fill energy is adaptively determined for each subband according to the shape of the frequency spectrum for each speech frame.

Abstract: In a binary system for generating sound, digital signals representing an audio analog signal are stored in memory, a first voltage level is produced for a time t.sub.1 proportional to the value of a respective stored digital signal, and a second voltage level is produced for a time t.sub.2 =t.sub.c -t.sub.1, where t.sub.c is a fixed clock signal time interval. The first and second voltage levels are successively supplied to a speaker to operate the speaker in a type of frequency modulated mode to produce sound corresponding to the audio analog signal.

Abstract: A speech recognition system estimates a set of Poisson intensities for a spoken word, each intensity representing a respectively different word from a vocabulary of words. Each of the functions used to calculate these intensities has two variable parameter values. In a training mode, the system changes the values of the respective variable parameters to optimize the likelihood that the results predicted by the estimates correspond to the actual spoken words. These optimized parameter values are then used by the system, in an operational mode, to recognize spoken words.

Abstract: To program a communication device to respond to voice commands, an individual desiring to operate the communication device sends a message to a repository of voice recognition codebooks. The repository device(s) responds by transmitting the codebook of that individual to at least one communication device, which stores the codebook therein. Thereafter, the communication device may respond to the voice commands of that individual.

Abstract: A speaker verification system receives input speech from a speaker of unknown identity. The speech undergoes linear predictive coding (LPC) analysis and transformation to maximize separability between true speakers and impostors when compared to reference speech parameters which have been similarly transformed. The transformation incorporated a "inter-class" covariance matrix of successful impostors within a database.