Showing papers on "String (computer science) published in 1981"

A level building dynamic time warping algorithm for connected word recognition

Abstract: Dynamic time warping has been shown to be an effective method of handling variations in the time scale of polysyllabic words spoken in isolation. This class of techniques has recently been applied to connected word recognition with high degrees of success. In this paper a level building technique is proposed for optimally time aligning a sequence of connected words with a sequence of isolated word reference patterns. The resulting algorithm, which has been found to be a special case of an algorithm previously described by Bahl and Jelinek, is shown to be significantly more efficient than the one recently proposed by Sakoe for connected word recognition, while maintaining the same accuracy in estimating the best possible matching string. An analysis of the level building method shows that it can be obtained as a modification to the Sakoe method by reversing the order of minimizations in the two-pass technique with some subsequent processing. This level building algorithm has a number of implementation parameters that can be used to control the efficiency of the method, as well as its accuracy. The nature of these parameters is discussed in this paper. In a companion paper we discuss the application of this level building time warping method to a connected digit recognition problem.

A theory of order relations in perceptual matching.

Abstract: A theory of order relations in the perceptual matching task relates order manipulations to research on retrieval processes and the representation of order information in memory. In experimental tests of the theory, presentation of a study string of letters to the subject was followed by a test string to which the subject responded same or different. The data of main interest concern the case where the test string is a permutation of the study string. When adjacent letters are switched, reaction time is long and accuracy low, suggesting that, in the comparison process, a test letter is not simply compared to the letter in the same position in the study string; rather, the comparison is distributed across positions. The memory model assumes that the representation of a letter is distributed (spread) over position and that the comparison process assesses the amount of overlap between the test string and the memory representation. The amount of overlap is transformed by a power function into the drift rate in a diffusion (random walk) comparison process. The diffusion retrieval model and overlap memory model are fitted to the data and goodness-of-fit is assessed. Shortcomings of alternative models are considered and applications of the model to related matching tasks are described.

Connected digit recognition using a level-building DTW algorithm

Abstract: In this paper we present a novel method for recognizing a string of connected digits based upon the use of a recently proposed level-building dynamic time warping (DTW) algorithm. The recognition system attempts to build up the string, level-by-level (i.e., digit-by-digit), by comparing portions of the test string to isolated digit reference patterns. A backtracking procedure is used to find the "best" string (i.e., minimum accumulated distance) as well as a set of reasonable alternative candidates. The system was tested on a number of talkers speaking variable length digit strings (from two to five digits) over dialed up telephone lines. String error rates of 4.8 percent and 4.6 percent were obtained for speaker-trained and speaker-independent systems. Word error rates of 0.7 percent (for speaker-trained tests) and 0.9 percent (for speaker-independant tests) were obtained. The digit reference templates were obtained from autocorrelation averaging of a pair of isolated word templates for each digit of the speaker-trained system, and from a clustering analysis of isolated words for the speaker-independent system.

Continuous speech recognition

Abstract: An improved speech recognition method and apparatus for recognizing keywords in a continuous audio signal are disclosed. The keywords, generally either a word or a string of words, are each represented by an element template defined by a plurality of target patterns. Each target pattern is represented by a plurality of statistics describing the expected behavior of a group of spectra selected from plural short-term spectra generated by processing of the incoming audio. The incoming audio spectra are processed to enhance the separation between the spectral pattern classes during later analysis. The processed audio spectra are grouped into multi-frame spectral patterns and are compared, using likelihood statistics, with the target patterns of the element templates. Each multi-frame pattern is forced to contribute to each of a plurality of pattern scores as represented by the element templates. The method and apparatus use speaker independent word models during the training stage to generate, automatically, improved target patterns. The apparatus and method further employ grammatical syntax during the training stage for identifying the beginning and ending boundaries of unknown keywords. Recognition is further improved by use of a plurality of templates representing "silence" or non-speech signals, for example, hum. Also, memory and computation load is reduced by use of modified (collapsed or folded) syntax flow graph logic, implemented by additional (augment) control numbers. A concatenation technique is employed, using dynamic programming techniques, to determine the correct identity of the word string.

Automatic tuning device

Abstract: A stringed musical instrument is provided with an automatic self-tuning device which tightens or loosens those of the instrument's strings which may be out of tune. Each string is provided with an independent automatic tuning device and all of the devices are operated simultaneously on demand to instantly and automatically reset the tension of the strings as required. Each of the devices includes means for sensing the tension of its associated string and means for comparing the sensed tension with a reference tension corresponding to the desired tuning for the string. Means are provided for increasing or decreasing the string tension in response to the nature and magnitude of the compared tensions. In an alternate version, string tension is sensed and a visual indication is given of an out-of-tune condition. The instrument is tuned in this embodiment manually.

String Matching in Real Time

Abstract: A sufficient condition for an on-line algorithm to be transformed into a real-time algorithm is given. This condition is used to construct real-time algorithms for various string-matching problems by random access machines and by Turing machines.

Method and system for matching encoded characters

Abstract: A means and method for comparing an incoming sequential string of digitally encoded characters from a database stored in a conventional memory against a pattern with an arbitrary number of elements, comprising specified characters or character types (alphabetic, numeric, delimiter, etc.) or tokens to indicate the matching of a specified or arbitrary number of input characters, is disclosed. The system comprises a number of digital machines, sequenced by control words fetched from their memories. The control words may indicate the current input character or character type of interest for each machine, the address of the potential next control word of the machine, a flag indicating the successful completion of a match, and other control fields. If the input character matches the character or type of interest, the machine's next control word will be that specified by the current control word, and optionally the next control word of one or more of the other machines will be forced to an address specified in the current control word. By properly specifying the control words in each machine the input character string can be compared against an arbitrary number of pattern elements, limited only by the ability to map the elements into the control word memories of the available machines.

An On-Chip Compare/Steer Bubble Sorter

Abstract: Two generic record-permutation bubble devices—the bubble ladder and the bubble string comparator—have been reported in the literature but not yet implemented. The former relies on the extensive use of external control lines, while the latter relies solely on the interaction between bubbles. The ladder has evolved into an odd- even sorter and then a rebound sorter, but, uufortunately, it is operated by a large number of control lines. This paper shows that equally efficient but more versatile sorters can be constructed from the bubble string comparators without the control lines. Moreover, the new sorter—an up-down sorter—will be implemented in the recently invented high-density, high-speed, coil-less perforated-sheet bubble devices.

Apparatus and method for recognizing spoken words

Abstract: A speech recognition technique is disclosed for recognizing words that are spoken at speeds that approach the speed of continuous speech. The characteristic form of each vocabulary word spoken by a user is stored in a reference array (110). The process of input speech recognition then includes the steps of. detecting pauses in the speech so as to define strings of words or syllables, and segments within each string: extracting (60) features from the input speech and using them to compare each segment and each combination of adjacent segments with the stored words (130) to obtain the best match; and finally determining (140) the optimum sequence of vocabulary words for each word string.

Generators in Icon

Abstract: Icon is a new programming language that includes a goal-directed expression evaluation mechanism. This mechanism is based on generators--expressions that are capable of producing more than one value. If the value produced by a generator does not lead to a successful result, the generator is automatically activated for an alternate value. Generators form an integral part of Icon and can be used anywhere. In addition, they form the basis for the string scanning facility and subsume some of the control expressions found in other languages. Several examples are given.

Phonological and orthographic factors in the word-superiority effect

Abstract: Phonological and orthographic aspects of a letter string were found to affect the identification of a component letter in three experiments. All involved a fixed set of target vowels presented in a fixed position in letter strings. Manipulations of the phonological nature of the target or the orthographic character of the string were made by adding a letter with the postexposure mask to the original CVC trigram. In Experiment 1, the addition of an E with the mask as a final letter to the string changed the pronunciation of the target vowel, whereas the addition of an S did not. Identification accuracy was higher with the S mask. In Experiment 2, either E or D could be added to CVCs that were equally orthographic but differentially pronounceable. The same added letter had quite different effects on accuracy, depending on its effect on target pronunciation and the orthographic regularity of the string. In Experiment 3, performance on targets in orthographic CVCs was lowered to the level of nonorthographic CVCs by adding a letter that rendered the entire string nonorthographic. The results are explained by assuming that phonological and graphemic codes are developed simultaneously but maintained in a nonindependent manner.

Continuous speech pattern recognizer

Abstract: This speech recognizer concatenates a string of reference isolated-words for comparison with the unknown string of connected-words. The invention includes a level-building (LB) algorithm, "level" implying a location in a sequence of words. A constrained endpoint dynamic-time-warp algorithm, in which the slope of the warping function is restricted between 1/2 and 2, is used to find the best alignment between an unknown continuous-word test pattern, and a concatenated sequence of L reference patterns. Properties of the LB algorithm include: modification of the references; back-track decision logic; heuristic selection of multiple candidates, and syntax constraints. As a result, the processing required is less than two-level dynamic-program-matching and sampling algorithms.

The string merging problem

Abstract: The string merging problem is to determine a merged string from a given set of strings. The distinguishing property of a solution is that the total cost of editing all of the given strings into this solution is minimal. Necessary and sufficient conditions are presented for the case where this solution matches the solution to the string-to-string correction problem. A special case where deletion is the only allowed edition operation is shown to have the longest common subsequence of the strings as its solution.

Japanese text inputting system having interactive mnemonic mode and display choice mode

Abstract: A text inputting device includes a keyboard for inputting a string of characters of a first kind by keying keys and a first device for displaying characters of a second kind corresponding to the character string supplied from the keyboard as candidate characters and selecting a desired character of the second kind from the candidate characters for supplying it to a processor. A second device is also provided for reading a character of the second kind corresponding to a mnemonic code in the character string supplied from the keyboard for supplying it to the processor. In order to determine whether the first or second device is used, a selection switch is provided for selectively activating the first or the second device when a character of the second type is to be inputted. Further, to assist in teaching the mnemonic code, when the first device is selected, either visual or oral correspondence is provided between the selected candidate character and its corresponding mnemonic code. The invention is particularly useful in inputting complex texts such as Japanese or other language texts.

Automatic labelling of continuous speech with a given phonetic transcription using dynamic programming algorithms

Abstract: A system is described which allows the mapping of a phonetic transcription onto an acoustic parameter representation of continuous speech. Linear prediction analysis, segmentation and formant tracking provide the acoustic parameters on a 5 ms time frame basis and a sequence of voiced, unvoiced and silent segments. The given phonetic transcription is expanded to include implicit phone sequences and transitions. Labelling is then performed in two stages. Segment labelling maps substrings of the expanded phone string onto the acoustic segments using a dynamic programming algorithm. The acoustic and phonetic units are correlated directly by means of a table of acoustic-phonetic rules. Frame labelling labels each time frame with a single phone using another dynamic programming algorithm based on the derivatives of energy and formant functions. The method is found to objectify and considerably facilitate the establishment of a time-locked acoustic-phonetic database.

Continuous speech recognition system

Abstract: Recognition of continuous speech by comparison with prestored isolated words may be confused by the merging together of spoken adjacent words (coarticulation). Improved recognition is attained by generating overlap-words, e.g., words whose first phoneme is the end phoneme of the preceding word in a string of words. The reference candidate series of overlap-words is transformed under dynamic time warping so as to time-match the utterance series of overlap-words.

Method and means for pipeline decoding of the high to low order pairwise combined digits of a decodable set of relatively shifted finite number of strings

Abstract: An apparatus for ensuring continuous flow through a pipeline processor as it relates to the serial decoding of FIFO Rissanen/Langdon arithmetic string code of binary sources. The pipeline decoder includes a processor (11, 23) and a finite state machine (21, FSM) in interactive signal relation. The processor generates output binary source signals (18), status signals (WASMPS, 31) and K component/K candidate next integer-valued control parameters (L0, k0; L1, k1; 25). These signals and parameters are generated in response to the concurrent application of one bit from successive arithmetic code bits, a K component present integer-value control parameter (52) and K component vector representation (T, TA) of the present internal state (51) of the associated finite state machine (FSM). The FSM makes a K-way selection from K candidate next internal states and K candidate next control parameters. This selection uses no more than K2 +K computations. The selected signals are then applied to the processor in a predetermined displaced time relation to the present signals in the processor. As a consequence, this system takes advantage of the multi-state or "memory" capability of an FSM in order to control the inter-symbol influence and facilitate synchronous multi-stage pipeline decoding.

Polyfunction programmable data receiver

Abstract: A polyfunction programmable data receiver comprising means for receiving a string of serial data followed by a control signal; means for detecting said signal comprising a pulse length analyzer, and a counter and a comparator; bidirectional means for setting up a digital word to be identified or for providing in parallel the data of said string according to the status of a mode selection circuit; means for providing an output signal if the comparison between said serial string and said set up digital word is favorable when the receiver operates as identifier, or an enabling signal when the receiver operates as a series/parallel converter. Some applications of this circuit are also disclosed.

On the line breaking problem in text formatting

Abstract: A basic problem in text formatting is that of determining the break points for separating a string of words into lines to obtain a formatted paragraph. When formatted text is required to be aligned with both the left and right margins, the choice of break points greatly affects the quality of the formatted document. This paper presents and discusses solutions to the line breaking problem. These include the usual line-by-line method, a dynamic programming approach, and a new algorithm which is optimal and runs almost as fast as the line-by-line method.

Boolean filtering method and apparatus

Abstract: A filtering method and apparatus for discriminating targets from noise in space-time data consisting of a plurality of binary data frames, each containing an equal number of points, are disclosed. Noise filtering for target detection is performed by recursively determining whether the points of a sequence of M of the binary data frames form at least one continuous string, or at the next highest level of discrimination, monotonic continuous string by logically comparing each binary data frame with the two binary data frames which precede and follow it on a point by point basis.

A generative model of working set dynamics

Abstract: An algorithm for generating a page reference string which exhibits a given working set size behavior in the time domain is presented, and the possible applications of such a string are discussed. The correctness of the algorithm is proved, and its computational complexity found to be linear in the length of the string. A program implementing the algorithm, which is performed in one pass and requires very little space, is briefly described, and some experimental results are given.

Digital Simulation Of Reticle Systems

Abstract: Various methods for simulation of reticle systems will be described which have practical applications for the simulation of missile guidance systems. The reticle pattern, as well as the received image, is digitized using common picture processing equipment. The sampled representations of the reticle and image are converted to a polar coordinate system, and the data is then put into a vector string. Setting up a cyclic matrix with this vector string, one can describe the periodic system signal by cyclic convolution of reticle string and image string. A similarity transform to the Jordan form of a quadratic cyclic matrix, by means of the discrete Fourier matrix, reduces the amount of calculations required. This is a great advantage in digital simulation. For this linear discrete system approximation, the transfer function in the Z-domain can be deduced. This approach gives a better insight into the overall performance of the control system. The analysis also shows the limitations of reticles, which has resulted in the decline of this technique for guided missile applications.

Parallel random access machines with powerful instruction sets

Abstract: A random access machine model that has capabilities for parallel processing and string manipulation is introduced. It is shown that NP is equal to the class of sets accepted by this model in nondeterministic timeO(logn), that PSPACE is equal to the class of sets accepted by this model in deterministic polynomial time and thatP is equal to the class of sets accepted by a restricted version of this model inO(logn) space. These results generalize to arbitrary time and storage bounds.