Showing papers on "Word error rate published in 1974"

On a New Class of Bounds on Bayes Risk in Multihypothesis Pattern Recognition

Abstract: An important measure concerning the use of statistical decision schemes is the error probability associated with the decision rule. Several methods giving bounds on the error probability are presently available, but, most often, the bounds are loose. Those methods generally make use of so-cailed distances between statistical distributions. In this paper a new distance is proposed which permits tighter bounds to be set on the error probability of the Bayesian decision rule and which is shown to be closely related to several certainty or separability measures. Among these are the nearest neighbor error rate and the average conditional quadratic entropy of Vajda. Moreover, our distance bears much resemblance to the information theoretic concept of equivocation. This relationship is discussed. Comparison is made between the bounds on the Bayes risk obtained with the Bhattacharyya coefficient, the equivocation, and the new measure which we have named the Bayesian distance.

Error-rate monitoring unit in a communication system

Abstract: A method and apparatus are disclosed for monitoring the accuracy of electronic data transmitted over a communication channel or telemetry link. An error code is generated for a unit of data transmitted and one for a unit of data received. The error code for a unit of data received is transmitted to the data transmitting station and then compared with the error code representing the unit of data transmitted. Means are provided to synchronize the comparison of error codes so that the two error codes compared represent the same data. After synchronization the bit error rate (BER) is calculated based on the number of miscompares that result for a given quantity of data transmitted. When the system exceeds a predetermined level corrective action is taken.

Speaker‐independent digit recognition system

Abstract: This paper describes an implementation of a speaker‐independent digit recognition system. The digit classification scheme is based on segmenting the unknown word into three regions and then making categorical judgments as to which of six broad acoustic classes each segment falls into. The measurements made on the speech waveform include energy, zero crossings, and two‐pole LPC analysis. A formal evaluation of the systems showed an error rate of 2.7% for a carefully controlled recording environment, and a 5.6% error rate for online recordings in a noisy computer room.

A practical application of a real-time isolated-word recognition system using syntactic constraints

Abstract: In this paper we describe an application of a real-time isolated-word recognition system, to the recognition of sentences of a language used in numerical command of machine tools. The acoustic level operates with dynamic matching procedure and knowledge about syntactics and semantics of the language is used to predict the incoming words. With such a syntax-directed system, real-time recognition of sentences pronounced word-by-word is very accurately achieved, even for several speakers. This system will then be adapted to the recognition of connected speech, as well as in the design of new languages for man-machine voice communication.

Decision-Aided Tracking Loops for Channels with Phase Jitter and Intersymbol Interference

Abstract: The theory of decision-aided carrier tracking loops as developed by Lindsey and Simon is extended to include the effects of intersymbol interference. A major theoretical result is the solution of the Fokker-Planck equation for the probability density function (pdf) of the phase error for a first-order tracking loop. This function is used to average upper and lower bounds on the demodulator error rate so that an estimate of the overall system error rate may be obtained. The modulation technique analyzed is double-sideband-pulse-amplitude pulse-amplitude modulation. Curves are presented which display the threshold performance of the system error rate at low modulation index.

Cepstral analysis technique for automatic speaker verification

Abstract: This paper describes new techniques for automatic speaker verification using telephone speech. The operation of the system is based on a set of functions of time obtained from acoustic analysis of a fixed, sentence-long utterance. Cepstrum coefficients are extracted by means of LPC analysis successively throughout an utterance to form time functions, and frequency response distortions introduced by transmission systems are removed. The time functions are expanded by orthogonal polynomial representations and, after a feature selection procedure, brought into time registration with stored reference functions to calculate the overall distance. This is accomplished by a new time warping method using a dynamic programming technique. A decision is made to accept or reject an identity claim, based on the overall distance. Reference functions and decision thresholds are updated for each customer. Several sets of experimental utterances were used for the evaluation of the system, which include male and female utterances recorded over a conventional telephone connection. Male utterances processed by ADPCM and LPC coding systems were used together with unprocessed utterances. Results of the experiment indicate that verification error rate of one percent or less can be obtained even if the reference and test utterances are subjected to different transmission conditions.

Simulation Results for the Decision-Directed MAP Receiver for M-ary Signals in Mulitiplicative and Additive Gaussian Noise

Abstract: Simulation results are presented for the error-rate performance of the recursive digital maximum a posteriori probability (MAP) detector for known M -ary signals in multiplicative and additive Gaussian noise. The structure of the digital simulation of the optimum detector is generally described, with specific results obtained for a quaternary signal and 2500 digit per second transmission rate. The simulation is focused on the aeronautical multipath communication problem. Plots of detection error rate versus additive signal-to-noise ratio are given, with the power ratio of multiplicative process to desired signal as a parameter. Results are presented for the cases where the detector has perfect knowledge of the first- and second-order statistics of the multiplicative and additive processes and also where these statistics are estimated in near real time. For comparison, the error rates of conventional coherent and noncoherent digital MAP detectors are also obtained. It is shown that with nonzero multiplicative noise, the error rates of the conventional detectors saturate at a level that is irreducible for increasing additive signal-to-noise ratio. The error rate of the optimum detector having perfect statistical knowledge continues to decrease rapidly with increasing additive signal-to-noise ratio. In the absence of multiplicative noise, the conventional coherent detector and the optimum detector are shown to exhibit identical performance. Suboptimum detectors, having less than perfect statistical knowledge, yield error rates bounded below by the optimum detector rates and bounded above by the conventional detector rates.

Note on Experimental Set and Word Recognition

Abstract: Two methods for studying the effect of experimental set on word recognition were contrasted. After study of a set of incomplete sentences, thresholds for words which would complete the sentences were not significantly lower than for inappropriate words. However, a significant difference was found between such words when study of incomplete sentences alternated with threshold determinations.