Parallel Dictionary Learning for Voice Conversion Using Discriminative Graph-Embedded Non-Negative Matrix Factorization

Aihara, Ryo; Takiguchi, Tetsuya; Ariki, Yasuo

doi:10.21437/interspeech.2016-227

Cited by 7 publications

(13 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wu et al [6] improved the original sparse representation by using both high-resolution and low-resolution features to capture spectral details and enforce temporal continuity. Aihara et al [8,9] and Berrak Sisman et al [12] incorporated phoneme information to solve the phoneme-mismatch problem. Liberatore et al [11,17] addressed the same issue by constructing compact exemplar dictionaries with a single centroid per phoneme.…”

Section: Literature Reviewmentioning

confidence: 99%

“…As described before, the standard objective function in eq. (2) can lead to phoneme mismatches [8,9]. To address this issue, we propose a Phoneme-Selective Objective Function (PSOF) based on the , norm [10].…”

Section: Promoting Phoneme Selectivity In Exemplar-based Vcmentioning

confidence: 99%

“…Mismatch Severity. Given that phoneme mismatches reduce the similarity between source and target weights [8,9], we use the dissimilarity between the source and target sparse representations of time-aligned parallel utterances to measure the mismatch severity, defined as,…”

Section: Effectiveness Of Phoneme-selective Objective Functionmentioning

confidence: 99%

“…However, exemplarbased methods lead to phoneme mismatches since the phoneme labels of the selected exemplars may not match the phonetic content of the frame being represented. These phoneme mismatches tend to be speaker-dependent, which reduces the similarity between the source and target sparse representations [8,9] and introduces distortions in the converted speech.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, this phoneme-mismatch problem becomes more severe as the size of the dictionary increases [8,9].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Improving Sparse Representations in Exemplar-Based Voice Conversion with a Phoneme-Selective Objective Function

et al. 2018

View full text Add to dashboard Cite

The acoustic quality of exemplar-based voice conversion (VC) degrades whenever the phoneme labels of the selected exemplars do not match the phonetic content of the frame being represented. To address this issue, we propose a Phoneme-Selective Objective Function (PSOF) that promotes a sparse representation of each speech frame with exemplars from a few phoneme classes. Namely, PSOF enforces group sparsity on the representation, where each group corresponds to a phoneme class. The sparse representation for exemplars within a phoneme class tends to activate or suppress simultaneously using the proposed objective function. We conducted two sets of experiments on the ARCTIC corpus to evaluate the proposed method. First, we evaluated the ability of PSOF to reduce phoneme mismatches. Then, we assessed its performance on a VC task and compared it against three baseline methods from previous studies. Results from objective measurements and subjective listening tests show that the proposed method effectively reduces phoneme mismatches and significantly improves VC acoustic quality while retaining the voice identity of the target speaker.

show abstract