VTLN-based cross-language voice conversion

2010

This paper presents the results on an Automatic Speech Recognition (ASR) framework that takes advantage of robust vocal tract length estimation methods for improving the performance of speech recognition in the presence of speakers with different conditions in age and gender. Well known techniques for Vocal Tract Length Normalization (VTLN) usually require previous stages for the estimation of the best warping factor for a given speaker, either by Maximum Likelihood (ML) estimates or by the calculation of acoustic features from the speakers like formant frecuencies through several utterances. This paper will show how to use robust framewise estimations of the vocal tract length to obtain a speaker dependent warping factor for achieving major improvements over all conditions of the TIDigits database. In the end, an updating function will be used to calculate an on-line estimate of the vocal tract length and the warping factor to use real time VTLN in speech recognition with similar results to the off-line strategies.

2006

In the framework of the European speech-to-speech translation project TC-Star, one of the research tasks is cross-language voice conversion. In the recent second evaluation campaign, five participants presented their intra-lingual as well as cross-language voice conversion systems that were applied to three languages. In this paper, we discuss the results of Siemens' submissions and describe the underlying system characteristics.

Cross-lingual speaker adaptation (CLSA) has emerged as a new challenge in statistical parametric speech synthesis, with specific application to speech-to-speech translation. Recent research has shown that reasonable speaker similarity can be achieved in CLSA using maximum likelihood linear transformation of model parameters, but this method also has weaknesses due to the inherent mismatch caused by differing phonetic inventories of languages. In this paper, we propose that fast and effective CLSA can be made using vocal tract length normalization (VTLN), where strong constraints of the vocal tract warping function may actually help to avoid the most severe effects of the aforementioned mismatch. VTLN has a single parameter that warps spectrum. Using shifted or adapted pitch, VTLN can still achieve reasonable speaker similarity. We present our approach, VTLN-based CLSA, and evaluation results that support our proposal under the limitation that the voice identity and speaking style of a target speaker don't diverge too far from that of the average voice model.

2002

Cross-language voice conversion is useful for many applications, and we are trying to apply the technique to a language training system for reducing voice individuality differences. In this paper, we describe experiments that test effectiveness of an extension of single-language voice conversion, to include cross-language utter- ances. The performance was investigated by objective and percep- tual evaluation using bilingual-speakers data for training. Then, the correlations between a computed distance measure and a hu- man perceptual pronunciation evaluation score were compared be- fore and after applying conversion. From these results, it was found that the cross-language voice conversion reduces speakers' voice differences between the pairs, and the phoneme based mea- sures show somewhat clearer correspondences to the human per- ceptual score in vowels' test after applying voice conversion.

Interspeech 2007, 2007

Most of the existing voice conversion methods calculate the optimal transformation function from a given set of paired acoustic vectors of the source and target speakers. The alignment of the phonetically equivalent source and target frames is problematic when the training corpus available is not parallel, although this is the most realistic situation. The alignment task is even more difficult in cross-lingual applications because the phoneme sets may be different in the involved languages. In this paper, a new iterative alignment method based on acoustic distances is proposed. The method is shown to be suitable for text-independent and cross-lingual voice conversion, and the conversion scores obtained in our evaluation experiments are not far from the performance achieved by using parallel training corpora.

IEEE Transactions on Audio, Speech, and Language Processing, 2013

This paper proposes a novel feature-space VTLN (vocal tract length normalization) method that models frequency warping as a linear interpolation of contiguous Mel filter-bank energies. The presented technique aims to reduce the distortion in the Mel filter-bank energy estimation due to the harmonic composition of voiced speech intervals and DFT (discrete Fourier transform) sampling when the central frequency of band-pass filters is shifted. This paper also proposes an analytical maximum likelihood (ML) method to estimate the optimal warping factor in the cepstral space. The presented interpolated filter-bank energy-based VTLN leads to relative reductions in WER (word error rate) as high as 11.2% and 7.6% when compared with the baseline system and standard VTLN, respectively, in a medium-vocabulary continuous speech recognition task. Also, the proposed VTLN scheme can provide significant reductions in WER when compared with state-of-the-art VTLN methods based on linear transforms in the cepstral feature-space. The warping factor estimated with the proposed VTLN approach shows more dependence on the speaker and more independence of the acoustic-phonetic content than the warping factor resulting from standard and state-of-the-art VTLN methods. Finally, the analytical ML-based optimization scheme presented here achieves almost the same reductions in WER as the ML grid search version of the technique with a computational load 20 times lower.

Ninth Annual Conference of the …, 2008

In this paper, we propose a Region-based multi-parametric Vocal Tract Length Normalization (R-VTLN) algorithm for the problem of automatic speech recognition (ASR). The proposed algorithm extends the well-established mono-parametric utterance-based VTLN algorithm of Lee and Rose [1] by dividing the speech frames of a test utterance into regions and by warping independently the features corresponding to each region using a maximum likelihood criterion. We propose two algorithms for classifying frames into regions: (i) an unsupervised clustering algorithm based on spectral distance, and (ii) an unsupervised algorithm assigning frames to regions based on phonetic-class labels obtained from the first recognition pass. We also investigate the ability of various mono-parametric and multiparametric warping functions to reduce the spectral distance between two speakers, as a function of phone. R-VTLN is shown to significantly outperform mono-parametric VTLN in terms of word accuracy for the AURORA4 database.

International Symposium on Multimedia, 2010

The cross lingual voice conversion problem refers to the replacement of a speaker's timbre or vocal identity in a recorded sentence, assuming that the source speaker and target speaker use different languages. This problem differs from typical voice conversion in the sense that the mapping of acoustical features cannot depend on time-aligned recordings of source and target speakers uttering the

IEEE Transactions on Speech and Audio Processing, 1998

Speech Communication, 2008

In this work, we present a speaker-normalization method based on the idea that the speaker-dependent scale-factor can be separated out as a fixed translation factor in an alternate domain. We also introduce a non-linear frequency-scaling model motivated by the analysis of speech data. The proposed shift-based normalization approach is implemented using a maximum-likelihood (ML) search for the translation factor in the alternate domain. The advantage of our approach is that we are able to show the relationship between conventional frequency-warping based vocal-tract length normalization (VTLN) methods and the methods based on shifts in psycho-acoustic scale thus providing a unifying framework for speaker-normalization. Additionally, in our approach it is simple to show that the shifting required for normalization can be expressed as a linear transformation in the cepstral domain. This is important for computational efficiency since we do not have to recompute the features by redoing the signal processing for each scale/translation factor as is usually done in conventional normalization. We present recognition results using our proposed approach on a digit recognition task and show that the non-linear scaling model provides relative improvement of 4% for adults and 7.5% for children when compared to the linear-scaling model.