Abstract: A speech synthesizer includes a first indicator for indicating the amplitude of a waveform by using a random number, a second indicator for indicating the superposition period for waveforms by using a random number, a waveform generator for generating first and second waveforms having an amplitude indicated by the first indicator, and a waveform superposition device for synthesizing an unvoiced speech waveform by superposing the second waveform generated by the waveform generator onto a waveform obtained by delaying the first waveform by a superposition period indicated by the second indication means. The speech synthesizer is capable of making the frequency characteristic of unvoiced speech analogous to that of white noise, and generating synthesized speech which is natural and analogous to an actual human voice.

Abstract: A pattern recognition system particularly useful for recognizing speech or handwriting. An input signal is first filtered by a filter bank having two stages where the outputs of the first stage is fed forward to the second stage of a significant number of filters and the output of the second stage is fed back to the first stage of a significant number of the filters. Such feedback enhances the signal-to-noise ratio and resembles the coupling between the different sections of the basilar membrane of the cochlear. The output of the filter bank is a two-dimensional frequency-time representation of the original signal. A second set of filters which takes as input two-dimensional signals, detects the presence of elementary tonotopic features such as the onset, rise, fall and frequency of any significant tones in a speech signal. A third set of filters detects any contrasts in the elementary features at various levels of resolution.

Abstract: In a method and circuit configuration for non-linear linkage of two binary words, a first method step is performed by shifting a first binary word to the right by a number of places associated with the first method step, and adding the unchanged first binary word thereto. N--2 further method steps are performed by shifting an outcome of addition of a preceding method step to the right by one place and by a number of places in respective method steps, and adding the unchanged first binary word thereto. A final method step is performed by shifting the outcome of the addition of the preceding method step to the right by one place and by a number of places associated with the final method step. A second binary word is split into n sub-binary words, and the numbers of places associated with the various method steps each correspond to one sub-binary word.

Abstract: A system for using sound to display data which includes the capability of storing, manipulating and retrieving data and data-to-sound parameter mappings for the purposes of controlling a sound generator with the data such that auditory reference beacons result. These beacons may be used to compare to sound resulting from the incoming data and/or to other beacons to orient a system user within a complex data set, and to enhance comprehension of system status and trends in the data. Incoming data to become the data component of the beacon generator is stored in memory, then, when recalled, is injected into a sonic map. The sonic map formats the data for control of the sound generator and routes it to selected parameters of a sound generator. By manipulating the beacon data and the sonic map, a flexible means of data inspection and reference are obtained.

Abstract: An apparatus and method for recognizing speech includes a memory for storing data representing a reference pattern composed of the combination of a word reference pattern and a silence pattern, and a calculator for calculating the differences between data representing the reference pattern and data representing input speech. The use of such a silence pattern in the reference pattern permits a word such as "other" to be distinguished from the word "mother".

Abstract: In a method of speech recognition, an input speech signal is analyzed. A result of the analysis is collated with first predetermined standard patterns. A result of the collation is outputted as a sequence of similarities. The sequence of the similarities is handled as feature parameters of the input speech signal, and the feature parameters are collated with second predetermined standard patterns of all recognition objects. A final speech recognition result is generated in accordance with a result of the collation between the feature parameters and the second predetermined standard patterns.

Abstract: An audio type signal is encoded. The signal is first divided into bands. For each band, a yardstick signal element is selected. The yardstick may be the signal element having the largest magnitude in the band, the second largest, closest to the median magnitude, or having some other selected magnitude. This magnitude is used for various purposes, including assigning bits to the different bands, and for establishing reconstruction levels within a band. The magnitude of non yardstick signal elements is also quantized. The encoded signal is also decoded. Apparatus for both encoding and decoding are also disclosed. The location of the yardstick element within its band may also be recorded and encoded, and used for efficiently allocating bits to non-yardstick signal elements. Split bands may be established, such that each split band includes a yardstick signal element and each full band includes a major and a minor yardstick signal element.

Abstract: A method and apparatus for editing the displayed voice wave form by marking the portion of interest on the screen is disclosed. Marked segment may then be deleted, for example, or copied into another segment in second voice editing window. In either case, pointers are established at the selected marker positions of the displayed voice segment and in the corresponding positions of uncompressed voice segments. The voice data is treated as a stream of fixed-length micro-segments, where there is a predictable correlation between the positions of the compressed and uncompressed data. In the implementation at hand, these micro-segments are 20 ms. in length. Editing is accomplished by modifying micro-segments in both the compressed and uncompressed segments simultaneously. When the user is satisfied with the result, the edited wave form is redrawn on the screen. The user may then SAVE the result, and the entire segment is rewritten to the data base, replacing the previous version.

Abstract: A speech dialogue system capable of realizing natural and smooth dialogue between the system and a human user, and easy maneuverability of the system. In the system, a semantic content of input speech from a user is understood and a semantic content determination of a response output is made according to the understood semantic content of the input speech. Then, a speech response and a visual response according to the determined response output are generated and outputted to the user. The dialogue between the system and the user is managed by controlling transitions between user states during which the input speech is to be entered and system states during which the system response is to be outputted. The understanding of a semantic content of input speech from a user is made by detecting keywords in the input speech, with the keywords to be detected in the input speech limited in advance, according to a state of a dialogue between the user and the system.

Abstract: A sound recording and reproducing apparatus includes a semiconductor memory, an acoustic signal writing unit for writing a digital acoustic signal to the semiconductor memory, and a read out unit for reading out digital acoustic signals addressed in the semiconductor memory during sound reproduction (replay). An address memory stores these addresses as the acoustic signals are read out. A silence detector detects recorded silent or nearly silent portions during sound reproduction. The address(es) of these silent portions is (are) sequentially stored. In response to a recall or replay instruction, a controller reads out from the semiconductor memory digital acoustic signals at memory addresses that follow those sequentially stored in the address memory. As a result, recorded sound information (as opposed to recorded silence) may be quickly accessed and replayed upon recall at the beginning of recorded acoustic signals.

Abstract: The present invention discloses a method for encoding and decoding human speech signals by generating a data base which stores a number of human speech signal types. The number of human speech signal types stored is sufficiently high enough to cover substantially all observable human speech. According to a first embodiment of the invention, a set of representative human speech signal curves is taken directly from natural human speech. According to a second embodiment of the invention, a predetermined set of speech signal parameters is used where maximum voice signal segment values are measured. According to a third embodiment of the invention, an adaptive set of speech signal parameters is used, where the encoder transmits a set of signal parameters to the decoder. Although the invention is specifically designed for human speech, it can also be used in connection with other audio signals, such as those of electronic musical instruments.

Abstract: A standard pattern comparing system recognizes a data by comparing the data with a standard pattern registered in a dictionary, a plurality of kinds of data being supplied to the standard pattern comparing system. The standard pattern comparing system has a plurality of dictionaries including a plurality of normal dictionaries and one master dictionary. All standard patterns stored in each normal dictionary and some of the standard patterns stored in the master dictionary corresponding to one kind of data. There being no duplicate standard pattern included in more than two dictionaries, and each standard pattern included in the master dictionary being commonly used for at least more than two normal dictionaries to correspond to an arbitrary kind of data. The present invention is especially useful for a voice recognition system in which a speaker dependent voice recognizer is used for a personal computer which manages a plurality of application programs.

Abstract: A signal processing card packaged on a bus of a personal computer has a bus master which is used to access the main memory of the personal computer. A large table of probability values required for speech recognition is held in the main memory. When a label to be processed is generated, only the necessary part of the table is read from the main memory to the memory on the signal processing card by direct memory access transfer to perform speech recognition processing.

Abstract: Disclosed is an invention concerning a quantization bit number allocating section for sub-band coding, wherein data important for human auditory sense are efficiently coded within a limited coding bit capacity to provide a high-quality digital audio signal. The quantization bit number allocating section comprises a level calculating section, a logarithm calculating section, an index calculating section, a quantization bit number calculating section, a logarithm weighting table, and a sub-band weighting table, wherein the quantization bit number is determined every prescribed time according to human auditory sense and the characteristic of an input digital audio signal.

Abstract: A word hypothesis module for speech decoding consists of four submodules: vowel center detection, bidirectional tree searches around each vowel center, forward-backward pruning, and additional short words hypotheses. By detecting the strong energy vowel centers, a vowel-centered lexicon tree can be placed at each vowel center and searches can be performed in both the left and right directions, where only simple phone models are used for fast acoustic match. A stage-wise forward-backward technique computes the word-beginning and word-ending likelihood scores over the generated half-word lattice for further pruning of the lattice. To avoid potential miss of short words with weak energy vowel centers, a lexicon tree is compiled for these words and tree searches are performed between each pair of adjacent vowel centers. The integration of the word hypothesizer with a top-down Viterbi beam search in continuous speech decoding provides two-pass decoding which significantly reduces computation time.

Abstract: A user inputs a network to an inputter of a network reformer for automatically reforming a network connecting basic units of data per a predetermined rewriting rule. The inputter sends the data to a partial path separator. The partial path separator separates the partial path matching the rewriting source before the predetermined rewriting rule is applied and sends it to a partial path reformer. The partial path reformer reforms a separated partial path by applying a predetermined rewriting rule to the separated partial path. A network merger receives a post-reform partial path and appropriately merges the partial path and the remaining network part, thereby creating a new network.

Abstract: A system is provided with a voice activated control apparatus which permits precise control of one or more system variables by means of voice commands uttered by the system operator. When a system variable is in a movement or change mode, the movement or change is terminated by any sound exceeding a preestablished acoustic threshold level. Any value of one or more system variables can be identified and appropriate data stored in a memory to permit the one or more system variables to return to an identified value or state with a single voice command. The control apparatus is combined with a screen monitor and/or an acoustic speaker to provide visible and/or acoustic responses to an operator. The control apparatus is practical for retrofitting existing remotely controllable systems.

Abstract: In a variable rate vocoder a method for determining a higher encoding rate of a set of encoding rates for unvoiced speech. The method is accomplished by generating an encoding rate indication based upon a first characteristic of an audio signal, determining a second characteristic of the audio signal, and modifying the encoding rate indication when the second characteristic of the audio signal is representative of unvoiced speech to provide a modified encoding rate indication corresponding to a higher encoding rate of the set of encoding rates.

Abstract: A speaker verification system which accepts or rejects the claimed identity of an individual based on analysis and measurements of the speaker's utterances. The utterances are elicited by prompting the individual seeking identification to read test phrases chosen at random by the verification system composed of words from a small vocabulary. Nearest-neighbor distances between speech frames derived from such spoken test phrases and speech frames of corresponding vocabulary "words" from previously stored utterances of the speaker seeking identification are computed along with distances between such spoken test phrases and corresponding vocabulary words for a set of reference speakers. The claim for identification is accepted or rejected based on the relationship among such distances and a predetermined threshold value.

Abstract: A method and system are provided for time aligning speech. Speech data is input representing speech signals from a speaker. An orthographic transcription is input including a plurality of words transcribed from the speech signals. A sentence model is generated indicating a selected order of the words in response to the orthographic transcription. In response to the orthographic transcription, word models are generated associated with respective ones of the words. The orthographic transcription is aligned with the speech data in response to the sentence model, to the word models and to the speech data.