Can we Automatically Transform Speech Recorded on Common Consumer Devices in Real-World Environments into Professional Production Quality Speech?—A Dataset, Insights, and Challenges

By Gautham J. Mysore. 2015

124 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Signal Processing

Citations in the corpus

High-Fidelity Audio Compression with Improved RVQGAN (2023, methodology): “For speech, we use the DAPS dataset [25], the clean speech segments from DNS Challenge 4 [10], the Common Voice dataset [2], and the VCTK dataset [39].”
Equalization matching of speech recordings in real-world environments (2016, methodology): “For our experiments, we used the data from the DAPS dataset [25].”
The Voice Conversion Challenge 2018: Promoting Development of Parallel and Nonparallel Methods (2018, methodology): “Like in the VCC 2016, the dataset used in the VCC 2018 is based on the device and production speech (DAPS) dataset [38], which includes recordings of professional US English speakers in a professional studio without significant noise effects and is available online for free(1).”
ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wild (2022, methodology): “In contrast to LA and PA evaluation data, both of which are derived solely from the VCTK database [10], the VCC 2018 and 2020 challenge data are derived from DAPS [33] and EMIME [34] corpora, respectively.”
The Voice Conversion Challenge 2016 (2016, methodology): “The dataset used in VCC 2016 is based on the DAPS (Data And Production Speech) dataset [38], which was recorded by professional US English speakers in a professional recording studio without signiﬁcant noise effects and is available online for free 1 .”
MOSNet: Deep Learning based Objective Assessment for Voice Conversion (2019, methodology): “The VCC 2018 corpus was prepared by selecting a part of speakers from the device and production speech (DAPS) dataset [12] recorded by professional US English speakers in a clean and noise-free environment.”
Chunked Autoregressive GAN for Conditional Waveform Synthesis (2021, methodology): “We use the pitch and periodicity representations described in Appendix A. Objective evaluation is performed on 256 randomly selected held-out examples from VCTK and DAPS.”
VocalSet: A Singing Voice Dataset (2018, background): “Recent generative audio models based on machine learning [11, 25] have mostly focused on speech applications, using multi-speaker speech datasets [6, 13].”
WavThruVec: Latent speech representation as intermediate features for neural speech synthesis (2022, methodology): “To evaluate whether it led to increased generalization capability to out-of-training speakers, we use the DAPS [34] as a test set of 10 male and 10 female unseen voices.”