Converting Foreign Accent Speech Without a Reference

Zhao, Guanlong; Ding, Shaojin; Gutierrez‐Osuna, Ricardo

doi:10.1109/taslp.2021.3060813

Cited by 11 publications

(12 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It converted the 80-dimensional Mel-spectrogram from the source-accented to the target-accented speech. The model followed the same implementation with the speech embedding to Mel-Spectrogram synthesizer in [3], and the only difference was the dimension of the input layer. In this system, converted speech does not preserve the speaker identity.…”

Section: A Experimental Systems and Setupsmentioning

confidence: 99%

“…So, we used the same speech data from speaker bdl and slt, and there were 4 target-for-source speaker pairs: bdl-for-ASI, slt-for-TNI, bdl-for-TXHC, and slt-for-LXC. The 256dimensional bottleneck features [3] extracted by a pretrained acoustic model were used as PPG inputs. The bottleneck details can be found in [29].…”

Section: A Experimental Systems and Setupsmentioning

confidence: 99%

“…The output was the 80dimensional Mel-spectrogram. The model configurations were the same as the speech synthesizer in [3].…”

Section: A Experimental Systems and Setupsmentioning

confidence: 99%

“…Ideally, it is expected to modify only the accent-related speech attributes, while preserving the speech content and the speaker identity. AC technique will enable a variety of real-life applications, such as language education [1], [2], customer service, movie dubbing [3], and personalized text-to-speech (TTS) [4].…”

Section: Introductionmentioning

confidence: 99%

“…Articulatory synthesizer methods take a different path by replacing the source pronunciation unit with the target counterpart through unit selection [6], [10], [11]. There are few recent works attempting to build a pronunciation correction model to transform the phonetic units from the source accent to the target's [3], [12]. However, these successful AC approaches require parallel speech data between the source and target accents during training, which limits the scope of applications [13], [14].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

TTS-Guided Training for Accent Conversion Without Parallel Data

Zhou

Zhang

et al. 2023

IEEE Signal Process. Lett.

View full text Add to dashboard Cite

Accent Conversion (AC) seeks to change the accent of speech from one (source) to another (target) while preserving the speech content and speaker identity. However, many existing AC approaches rely on source-target parallel speech data during training or reference speech at run-time. We propose a novel accent conversion framework without the need for either parallel data or reference speech. Specifically, a text-to-speech (TTS) system is first pretrained with target-accented speech data. This TTS model and its hidden representations are expected to be associated only with the target accent. Then, a speech encoder is trained to convert the accent of the speech under the supervision of the pretrained TTS model. In doing so, the source-accented speech and its corresponding transcription are forwarded to the speech encoder and the pretrained TTS, respectively. The output of the speech encoder is optimized to be the same as the text embedding in the TTS system. At run-time, the speech encoder is combined with the pretrained speech decoder to convert the source-accented speech toward the target. In the experiments, we converted English with two source accents (Chinese/Indian) to the target accent (American/British/Canadian). Both objective metrics and subjective listening tests successfully validate that the proposed approach generates speech samples that are close to the target accent with high speech quality.

show abstract

Section: A Experimental Systems and Setupsmentioning

confidence: 99%

Section: A Experimental Systems and Setupsmentioning

confidence: 99%

“…The output was the 80dimensional Mel-spectrogram. The model configurations were the same as the speech synthesizer in [3].…”

Section: A Experimental Systems and Setupsmentioning

confidence: 99%