B.S. Atal, "Automatic speaker recognition based on pitch contours," JASA, vol. 52, pp.1687-1697, 1972.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....tract by a time varying source (vibration of the vocal cords) The acoustic correlate of the vibration of the vocal cords is the fundamental frequency (F 0 ) or pitch frequency [12] In this paper, we de ne pitch to be F 0 . Pitch is a speaker speci c feature and is used for speaker recognition [13]. The voice source parameters include the type of phonation (voiced or unvoiced) and the measure of periodicity (F 0 ) of the speech signal, if it is voiced. Thus estimation of pitch implicitly provides information about voicing. The presence vs. absence of voicing plays a vital role in phonetics. ....

B. S. Atal, \Automatic speaker recognition based on pitch contours," JASA, vol. 52, no. 6, pp. 1687-1697, 1972.

....it possible to explore prosodic features for this study. The use of prosody for speaker ID is not a new idea. For example, SRI fielded a system incorporating prosodic features in the 1998 NIST speaker ID evaluation [6] and publications on prosody based speaker ID go back at least 30 years [7]. The novelty of our system lies in our access to an unusually wide variety of prosodic indicators via SRI s database, coupled with the availability of test and training data based on entire conversation sides. We contrast the results of our prosody based speaker ID approach with a system which ....

B. S. Atal, "Automatic speaker recognition based on pitch contours," JASA, vol. 52, pp. 1687--1697, 1972.

.... this approach also allows the recovery of the pitch period when using a telephone channel (which limits the bandwidth of speech signals to between 300 and 3400 Hz) Unfortunately the auto correlation method (and other time domain techniques, such as the Normalized Cross Correlation Method [9] and the Average Magnitude Di#erence Method [94, 119] su#er from pitch doubling and halving as well as other errors [56] If the signal is periodic with period P , it is also periodic with period 2P , 3P , etc. Hence, will also have maxima at lags equal to 2P , 3P , etc. Due to the presence ....

B. S. Atal, Automatic Speaker Recognition Based on Pitch Contours, PhD Thesis, Polytechnic Institute of Brooklyn, 1968.

.... this approach also allows the recovery of the pitch period when using a telephone channel (which limits the bandwidth of speech signals to between 300 and 3400 Hz) Unfortunately the auto correlation method (and other time domain techniques, such as the Normalized Cross Correlation Method [1] and the Average Magnitude Di erence Method [36, 50] su er from pitch doubling and halving as well as other errors [23] If the signal is periodic with period P , it is also periodic with period 2P , 3P , etc. Hence, fR(k)g k=0 will also have maxima at lags equal to 2P , 3P , etc. Due to the ....

B. S. Atal, Automatic Speaker Recognition Based on Pitch Contours, PhD Thesis, Polytechnic Institute of Brooklyn, 1968.

....by voice, in various services . Therefore the term speaker recognition in this context implies only the AM. That is why, throughout this work, the terms VM and AM are used to specifically refer to the particular type of SR. Practical SR experiments using the AM were undertaken in the early 70 s [36, 37], long before equivalent VM based systems were even considered. In the early 70 s researchers such as Furui et al. [36, 38] and Atal [37] performed some of the first AM experiments. These early experiments laid the foundations for future investigations into optimum feature sets, classifiers and ....

....the terms VM and AM are used to specifically refer to the particular type of SR. Practical SR experiments using the AM were undertaken in the early 70 s [36, 37] long before equivalent VM based systems were even considered. In the early 70 s researchers such as Furui et al. [36, 38] and Atal [37] performed some of the first AM experiments. These early experiments laid the foundations for future investigations into optimum feature sets, classifiers and databases. 37 Fortunately, the substantial increase in computer power over recent years, has meant that data intensive signal processing ....

B.S. Atal. Automatic Speaker Recognition Based on Pitch Contour. JASA, 52:1987-1697, 1972.

....coecients are obtained from the 16 predictor coecients using 20ms frames for every 10ms shift. Silence frames are removed using an amplitude threshold. The cepstral coecients characterize the spectrum of the speech signal which is primarily determined by the vocal cavities and the articulators [5] [2]. The spectral information of the same sound uttered by two persons di er due to the physical di erences between the vocal organs and the manner in which they use them during speech production [2] Since it is di cult to store the templates of all sounds spoken by the speaker, the distribution of ....

.... of the speech signal which is primarily determined by the vocal cavities and the articulators [5] 2] The spectral information of the same sound uttered by two persons di er due to the physical di erences between the vocal organs and the manner in which they use them during speech production [2]. Since it is di cult to store the templates of all sounds spoken by the speaker, the distribution of the speaker speci c features extracted from the speech signal is captured. In this context, the nonlinear models like AANN are used to capture the complex distribution of the speakerspeci c ....

B. S. Atal. Automatic speaker recognition based on pitch contours. J. Acoust. Soc. Amer., 52(6):1687{ 1697, 1972.

.... this approach also allows the recovery of the pitch period when using a telephone channel (which limits the bandwidth of speech signals to between 300 and 3400 Hz) Unfortunately the auto correlation method (and other time domain techniques, such as the Normalized Cross Correlation Method [9] and the Average Magnitude Difference Method [90, 115] suffer from pitch doubling and halving as well as other errors [53] If the signal is periodic with period P, it is also periodic with period 2P, 3P, etc. Hence, R(k) will also have maximas at lags equal to 2P, 3P, etc. Due to the ....

B. S. Atal, Automatic Speaker Recognition Based on Pitch Contours, PhD Thesis, Polytechnic Institute of Brooklyn, 1968.

.... retro ex vowels and nasals, voice pitch frequency, amplitude spectrum of vowels and nasals, slope of the glottal source spectrum, word duration and pitch frequency [6] 7] Prosodic features like mean F 0 [8] mean and variance of the pitch periods in voiced segments and energy contours of speech [9] have been used. The rst four statistics of the pitch, i.e. mean, variance, skew and kurtosis have also been used [10] In the traditional approach of indexing spoken documents, where speech is rst converted to text prior to indexing, cepstral coecients have been the predominant choice as ....

B. S. Atal, \Automatic speaker recognition based on pitch contours," JASA, vol. 52, no. 6, pp. 1687-1697, 1972.

....tract by a time varying source (vibration of the vocal cords) The acoustic correlate of the vibration of the vocal cords is the fundamental frequency (F 0 ) or pitch frequency [12] In this paper, we define pitch to be F 0 . Pitch is a speaker specific feature and is used for speaker recognition [13]. The voice source parameters include the type of phonation (voiced or unvoiced) and the measure of periodicity (F 0 ) of the speech signal, if it is voiced. Thus estimation of pitch implicitly provides information about voicing. The presence vs. absence of voicing plays a vital role in ....

B. S. Atal, "Automatic speaker recognition based on pitch contours," JASA, vol. 52, no. 6, pp. 1687-- 1697, 1972.

....tract by a time varying source (vibration of the vocal cords) The acoustic correlate of the vibration of the vocal cords is the fundamental frequency (F 0 ) or pitch frequency [5] In this paper, we de ne pitch to be F 0 . Pitch is a speakerspeci c feature and is used for speaker recognition [6]. In this work, by clustering the pitch we intend to cluster the speakers so as to derive appropriate models which could help in reducing the e ect of speaker variability on the system performance. The voice source parameters include the type of phonation (voiced or unvoiced) and the measure of ....

Bishnu S. Atal, \Automatic speaker recognition based on pitch contours," Journal of the Acoustical Society of America, vol. 52, no. 6, pp. 1687-1697, 1972.

....dependent on the shape and size of the vocal tract, which is unique to a speaker and the sound that is being produced. Fundamental frequency (F0) also carries speaker specific information, because F0 is dependent on accents, different phonological forms, behavior and other individualistic factors [42, 1]. To compute MFCCs, the speech signal was processed through a number of steps. First, the digitized utterances were initially passed through a pre emphasis filter, which enhances higher frequency components of the speech samples, and attenuates lower frequency components. Next, a short time ....

....frequency (F0) energy and duration. These features attempt to measure psychophysical perceptions of intonation, stress, and rhythm, which are presumably characteristics humans use to differentiate between speakers [6] Prosodic features have also proven to be robust in noisy environments [42, 17, 1]. Therefore, these features show great potential for the speaker verification task. To estimate F0, we used the ESPS tracker, in particular the FORMANT function [7] For each frame of sampled data, FORMANT estimates speech formant trajectories, fundamental frequency, and other related information. ....

B. Atal. Automatic speaker recognition based on pitch contours. JASA, 52:1687-- 1697, 1972.

....in the size of the vocal tract cavities produce differences in the spectrum of the speech signal. The length of the vocal tract affects the overall spectrum [4] ffl Variations in the size of the vocal folds are associated with changes in the average pitch or fundamental frequency of speech [3]. ffl Variations in velum and size of nasal cavities produce spectral differences in nasalized speech sounds [86, 64] ffl The configuration of the teeth and palate affects frication [64, 46] 17 ffl Behavioral traits such as speaking rate [64] breathing, nasalization and dialect affect the ....

B. S. Atal. Automatic speaker recognition based on pitch contours. Journal of the Acoustical Society of America, 52(6):1687--1697, December 1972.

....to be preferred in speaker recognition for this very reason. In general, though, features that are based on some kind of short term spectral estimate are used in speaker recognition much as they are in ASR. In addition, though, pitch information is sometimes used if it can be estimated reliably [1, 10]. Finally, the effects of transmission channel variability are usually reduced by using techniques initially proposed for speech recognition, such as cepstral mean subtraction or RASTA PLP. However, these techniques also can filter out important speaker specific characteristics. Further research ....

Atal, B.S., Automatic speaker recognition based on pitch contours, Journal of Acoustical Society of America, vol. 52, pp. 1687-1697, 1972.

....of speaker identification includes feature extraction and classification. With respect to feature extraction, many kinds of individual features covering from the characteristics of vocal cords to speech spectrum have already been investigated and turned out to be useful to speaker identification [3, 4, 18, 21, 22, 23, 24, 28, 34, 44, 60, 70]. Unfortunately, none of those features is perfect for robustness so that there is less agreement on which parameterization of the speech spectrum to use for features [18, 24, 28, 52, 49] In addition, some researchers intended to lump two or more features together into a composite feature [24, ....

....particular, the problem becomes quite serious when the techniques of neural computing with time delay [6, 8, 9, 13, 62] are used. On the other hand, several kinds of classifiers have been also applied in speaker identification [9, 18, 24, 28, 49, 63] These classifiers include distance classifiers [3, 4, 25, 33, 42], neural network classifiers [6, 7, 8, 11, 12, 13, 14, 19, 32, 46, 47, 54] and classifiers based upon parametric or non parametric density estimation [28, 29, 52, 57, 59] Since there are many kinds of features and classifiers, speaker identification becomes a typical task which needs to combine ....

[Article contains additional citation context not shown here]

B. S. Atal. Automatic speaker recognition based on pitch contour. J. Acoust. Soc. Am., 52(6):1687--1697, 1972.

No context found.

B.S. Atal, "Automatic speaker recognition based on pitch contours," JASA, vol. 52, pp.1687-1697, 1972.

No context found.

B. S. Atal, "Automatic speaker recognition based on pitch contours," J. Acoust. Soc. Amer., vol. 52, pp. 1687--1697, 1972.

No context found.

Atal, B. Automatic speaker recognition based on pitch contours. Journal of the Acoustic Society of America 52, 6 (1972), 1687--1697.

No context found.

B.S. Atal, "Automatic speaker recognition based on pitch contours," JASA, vol. 52, pp.1687-1697, 1972.

No context found.

B.S.Atal, "Automatic speaker recognition based on pitch contours," Journal of the Acoustical Society of America, vol. 52, no. 6, 1972.

No context found.

B. S. Atal, "Automatic speaker recognition based on pitch contours," J. Acoust. Soc. Amer., vol. 52, no. 6, pp. 1687--1697, 1972.

No context found.

B.S. Atal, "Automatic speaker recognition based on pitch contours," J. Acoust. Soc. Amer., 52 (6), 1687--1697, 1972.

No context found.

Atal B. S., Automatic speaker recognition based on pitch contours, In The Journal of the Acoustical Society of America, pp. 1687-1697, Vol. 52, 1972.

No context found.

B. S. Atal, Automatic Speaker Recognition Based on Pitch Contours, PhD Thesis, Polytechnic Institute of Brooklyn, 1968.

No context found.

B. S. Atal, Automatic Speaker Recognition Based on Pitch Contours, PhD Thesis, Polytechnic Institute of Brooklyn, 1968.

No context found.

B. S. Atal, "Automatic speaker recognition based on pitch contours," J. Acoust. Soc. Amer., vol. 52, pp. 1687--1697, 1972.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC