Showing papers on "Speaker recognition published in 1973"

Speaker identification by speech spectrograms: some further observations

Abstract: This letter reviews recent research on speaker identification by means of comparisons of speech spectrograms by human observers. Various factors affecting the reliability of identification are discussed, particularly those that would be present in practical forensic situations. Our interpretations of the new data lead us to reiterate our previous conclusion: that the degree of reliability of identification under practical conditions has not been scientifically established.

Speaker Recognition and Verification using Linear Prediction Analysis

Abstract: The speaker recognition and verification effectiveness of a set of 92 measurements was examined. These measurements included the formant structure of vowels, the duration of certain speech events, the dynamic behavior of the formant contours, various aspects of the pitch contour throughout an utterance, formant bandwidths, glottal source “poles,” and pole and zero locations during the production of nasals and strident consonants. Linear prediction methods were employed in the analysis, and a probability of error criterion was devised to evaluate the speaker‐characterizing potential of the measurements. The experimental speech data were collected during five different recording sessions, the last session being 312 years after the original recordings. The measurements that were found most useful were related to the nasals, certain vowel resonances, certain temporal attributes, and average fundamental frequency. A speaker identification experiment using only the five “best” measurements resulted in only one ...

Techniques for Objective Speaker Identification

Abstract: A technique has been developed and tested for identifying from voice recordings which, if any, of a number of possible speakers corresponds to a particular object speaker recording. Segments of the same set of selected phonemes by the object speaker and all suspect speakers are Fourier transformed. These phoneme spectra, and features derived therefrom, are utilized to make measures of voice similarity. Experiments with voice recordings by 140 men have yielded estimates of the probability of correct identification as a function of the number of suspects for both forced‐choice and open‐choice tests. Identification reliability has also been related to number of phonemes, number of utterances, and number and type of features. Open‐choice decision reliability of 99% or better for small numbers of suspects was demonstrated. Another decision procedure, whereby decisions are made in only 95% to 99% of cases, can reduce error probabilities essentially to zero.