L.M. Arslan, D.Talkin, Voice conversion by codebook mapping of line spectral frequences and excitation spectrum, Proceedings EUROSPEECH, 1997,3:1347- 1350.  Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
High Resolution Voice Transformation - Kain (2001) (2 citations) (Correct)
....on the magnitude spectrum [91] DFW is a technique that aims at obtaining an optimal, nonlinear warping function of the frequency axis to simulate the changes of speaker char acteristics. However, the authors found DFW to be inferior to a more traditional spectral envelope mapping. Arslan et al. [6, 7, 5] formulated a codebook based transformation of LPC residuals using a weighted combination of xcitation filters, one for each speech class of a spectral envelope transformation. The excitation filters were derived from the average source and target residual spectra within one class. This ....
.... (DTW) algorithm [74] in most of the previous approaches [2, 91, 61, 84, 52] However, it is also possible to use a form of linguistic labeling, as obtained from the states of an unsupervised hidden Markov model (HMM) 7, 5] by forced alignment speech recognition [36] or by a phonetic classifier [6, 5]. We now present three different methods for implementing a transformation function. Mapping codebooks One of the earliest works in the field of VT used a transformation technique called mapping code books [2] In this implementation, the codevectors of a source codebook have a one to one ....
[Article contains additional citation context not shown here]
AISLAN, L. M., AND TALKIN, D. Voice conversion by codebook mapping of line spectral frequencies and excitation spectrum. In Proceedings of Eurospeech '97 (Rhodes, Greece, September 1997), vol. 3, pp. 1347-1350.
The Impact Of Speech Recognition On Speech Synthesis - Ostendorf, Bulyko (2002) (1 citation) (Correct)
....approach [21] However, these techniques provide only a partial solution, since they fail to capture the prosodic idiosyncrasies and glottal source characteristics of the speaker. There is some work showing that one can apply the same adaptation techniques to both spectral and source parameters [3, 67], but these operate at essentially the frame level and much more research is needed to effectively adapt the suprasegmental characteristics of a speaker. 5. SPECULATIONS There is some concern that there has been an over emphasis on ever larger data sets and faster computing (consistent with ASR) ....
L. Arslan and D. Talkin, "Voice conversion by codebook mapping of line spectral frequencies and excitation spectrum," Proc. ICASSP, 3:1347-1350, 1997.
The Need for Increased Speech Synthesis Research: Report .. - Sproat, Ostendorf.. (Correct)
....in a another voice based on a small amount of adaptation data. Most previous approaches to this problem have relied simply on various forms of regression mapping between source target pairs of spectral features (e.g. LPC or cepstral coefficients) Kain and Macon, 1998, Stylianou et al. 1995, Arslan and Talkin, 1997] Although many acoustic correlates of voice identity are carried in the spectral envelope [Kuwabara and Sagisaka, 1995] a frame by frame spectral mapping does not capture systematic cross speaker variations in timing, voice quality, intonation, or other supra segmental features. Much further ....
Arslan, L. M. and Talkin, D. (1997). Voice conversion by codebook mapping of line spectral frequencies. In Proc. Eurospeech, pages 1347--1350.
Chapter 7 Summary and Future Work - Within The Last (Correct)
....every verification system remains open to attack by voice mimicry. The imposter, in this case, may attempt to gain access by manipulating his her own voice to take on the characteristics of the customer. In recent years, several authors have proposed methods for performing voice conversion [2, 7, 92, 124, 178]. These techniques require identical speech from both the source (imposter) and target (customer) voices to be available. In general, voice conversion methods attempt to learn a spectral mapping between the source and target voices (e.g. a statistical transformation or codebook mapping) ....
L. Arslan and D. Talkin. Voice conversion by codebook mapping of line spectral frequencies and excitation spectrum. In Proc. Eurospeech'97, volume 3, pages 1347--1350, Rhodes, Greece, September 1997.
Spectral Voice Conversion For Text-To-Speech Synthesis - Kain, Macon (1998) (3 citations) (Correct)
....bandwidth, which have been shown to be perceptually relevant for speaker identity. 4. Since the training cost function is the mean squared error, a bark scaling weights errors in accordance with the frequency sensitivity of human hearing. LSF features were recently applied to voice conversion in [2] as well. 3.3 Conversion To convert an utterance, the speech synthesis signal processing engine is modified as follows: spectral vectors drawn from the source speaker s database are converted using the conversion function F with parameters from the trained GMM. The pitch of the source speaker s ....
L. M. Arslan and David Talkin, "Voice conversion by codebook mapping of line spectral frequencies and excitation spectrum," Proceedings of EUROSPEECH, pp. 1347-1350, September 1997.
Subband Based Voice Conversion - Levent (2002) Self-citation (Arslan) (Correct)
....space. Previous research on voice conversion employing codebook based methods include [1] 2] and [3] The acoustic spaces of the source and target speaker are represented using acoustical features such as formant frequencies [4] LPC cepstrum coefficients, Line Spectral Frequencies (LSFs) [5] and harmonic plus noise model parameters [6] In this study, we investigate the use of Discrete Wavelet Transform (DWT) and subband processing as a modification to the Speaker Transformation Algorithm Using Segmental Codebooks (STASC) 1] as well as implementing a VCS system using both STASC and ....
Arslan, L.M., Talkin, D., "Voice Conversion by Codebook Mapping of Line Spectral Frequencies and Excitation Spectrum", Proceedings EUROSPEECH 1997, Rhodes, Greece, Vol. 3, pp. 1347-1350.
Speaker Transformation Algorithm using Segmental Codebooks (STASC) - Arslan (1999) (4 citations) Self-citation (Arslan) (Correct)
No context found.
L.M. Arslan and D. Talkin (1997). "Voice Conversion by Codebook Mapping of Line Spectral Frequencies and Excitation Spectrum". In Proceedings EUROSPEECH, volume 3, pp. 1347-- 1350, Rhodes, Greece.
Codebook Based Face Point Trajectory Synthesis Algorithm.. - Arslan, Talkin (1998) (1 citation) Self-citation (Arslan Talkin) (Correct)
No context found.
L.M. Arslan and D. Talkin (1997). "Voice Conversion by Codebook Mapping of Line Spectral Frequencies and Excitation Spectrum". In Proc. EUROSPEECH, Vol. 3, pp. 1347--1350, Rhodes, Greece.
Voice Morphing Using the Generative Topographic Mapping - Orphanidou, Moroz, Roberts (Correct)
No context found.
L.M. Arslan, D.Talkin, Voice conversion by codebook mapping of line spectral frequences and excitation spectrum, Proceedings EUROSPEECH, 1997,3:1347- 1350.
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