98 research outputs found
## SUPERSEDED: THIS DATASET HAS BEEN REPLACED. ## Noisy speech database for training speech enhancement algorithms and TTS models
No full text## SUPERSEDED: THIS DATASET HAS BEEN REPLACED by the one which can be found at https://doi.org/10.7488/ds/2117. ## Clean and noisy parallel speech database. The database was designed to train and test speech enhancement methods that operate at 48kHz. A more detailed description can be found in the paper associated with the database.
Some of the noises were obtained from the Demand database, available here: http://parole.loria.fr/DEMAND/
The speech database was obtained from the Voice Banking Corpus, available here: http://homepages.inf.ed.ac.uk/jyamagis/release/VCTK-Corpus.tar.gzThe files are wav format audio data sampled at 48kHz. Each file contains a sentence recorded by a range of speakers in quiet studio conditions. This audio material was added to a range of different noise signals, constituting the parallel noisy dataset. Accompanying each audio file there is a text file containing the orthographic transcription of what was said in that particular audio sample
Reverberant speech database for training speech dereverberation algorithms and TTS models
No full textReverberant speech database. The database was designed to train and test speech dereverberation methods that operate at 48kHz. Clean speech was made reverberant by convolving it with a room impulse response. The room impulse responses used to create this dataset were selected from: - The ACE challenge (http://www.commsp.ee.ic.ac.uk/~sap/projects/ace-challenge/); - The MIRD database (http://www.iks.rwth-aachen.de/en/research/tools-downloads/multichannel-impulse-response-database/); - The MARDY database (http://www.commsp.ee.ic.ac.uk/~sap/resources/mardy-multichannel-acoustic-reverberation-database-at-york-database/). The underlying clean speech data can be found in: https://doi.org/10.7488/ds/2117.The files are wav format audio data sampled at 48kHz. Each file contains a sentence recorded by a range of speakers in quiet studio conditions. This audio material was convolved with a range of different room impulse responses, constituting the parallel reverberant dataset. Accompanying each audio file there is a text file containing the orthographic transcription of what was said in that particular audio sample
Noisy reverberant speech database for training speech enhancement algorithms and TTS models
No full textNoisy reverberant speech database. The database was designed to train and test speech enhancement (noise suppression and dereverberation) methods that operate at 48kHz.
Clean speech was made reverberant and noisy by convolving it with a room impulse response and then adding it to a noisy signal that was also convolved with a room impulse response. The room impulse responses used to create this dataset were selected from: - The ACE challenge (http://www.commsp.ee.ic.ac.uk/~sap/projects/ace-challenge/)
- The MIRD database (http://www.iks.rwth-aachen.de/en/research/tools-downloads/multichannel-impulse-response-database/)
- The MARDY database (http://www.commsp.ee.ic.ac.uk/~sap/resources/mardy-multichannel-acoustic-reverberation-database-at-york-database/)
The underlying clean speech data can be found in: https://doi.org/10.7488/ds/2117 .
The speech-shaped and babble noise files that were used to create this dataset are available here: http://homepages.inf.ed.ac.uk/cvbotinh/se/noises/
Noisy speech database for training speech enhancement algorithms and TTS models
No full textClean and noisy parallel speech database. The database was designed to train and test speech enhancement methods that operate at 48kHz. A more detailed description can be found in the papers associated with the database. For the 28 speaker dataset, details can be found in: C. Valentini-Botinhao, X. Wang, S. Takaki & J. Yamagishi, "Speech Enhancement for a Noise-Robust Text-to-Speech Synthesis System using Deep Recurrent Neural Networks", In Proc. Interspeech 2016. For the 56 speaker dataset: C. Valentini-Botinhao, X. Wang, S. Takaki & J. Yamagishi, "Investigating RNN-based speech enhancement methods for noise-robust Text-to-Speech”, In Proc. SSW 2016. Some of the noises used to create the noisy speech were obtained from the Demand database, available here: http://parole.loria.fr/DEMAND/ . The speech database was obtained from the CSTR VCTK Corpus, available here: https://doi.org/10.7488/ds/1994. The speech-shaped and babble noise files that were used to create this dataset are available here: http://homepages.inf.ed.ac.uk/cvbotinh/se/noises/
Alba speech corpus
No full textSingle speaker read speech corpus of a Scottish accented female native English speaker (Alba). The corpus was recorded in four speaking styles: plain (normal read speech, around 4 hours of recordings), fast (speaking as fast as possible, around 20 mins), clear_c (computer directed speech, around 28 mins) and clear_h (hearing impaired directed speech, around 20mins). Audio files are sampled at 48kHz, segmented into sentences and saved in uncompressed formant (.wav). The underlying sentences are stored in separate text files (.txt)
Hurricane natural speech corpus - higher quality version
No full textSingle male native British-English talker recorded producing three speech sets (Harvard sentences, Modified Rhyme Test, news sentences) in quiet and while the talker was listening to speech-shaped noise at 84dB(A).
This is a higher quality alternative (48kHz sampling rate as oppose to 16kHz) of the dataset available in: https://doi.org/10.7488/ds/140 . Note however that sentence alignment is not the same, i.e. the amount of silence in the beginning and the end of each sentence is different in each dataset
Noisy speech database for training speech enhancement algorithms and TTS models
No full textClean and noisy parallel speech database. The database was designed to train and test speech enhancement methods that operate at 48kHz. A more detailed description can be found in the papers associated with the database. For the 28 speaker dataset, details can be found in: C. Valentini-Botinhao, X. Wang, S. Takaki & J. Yamagishi, "Speech Enhancement for a Noise-Robust Text-to-Speech Synthesis System using Deep Recurrent Neural Networks", In Proc. Interspeech 2016. For the 56 speaker dataset: C. Valentini-Botinhao, X. Wang, S. Takaki & J. Yamagishi, "Investigating RNN-based speech enhancement methods for noise-robust Text-to-Speech”, In Proc. SSW 2016. Some of the noises used to create the noisy speech were obtained from the Demand database, available here: http://parole.loria.fr/DEMAND/ . The speech database was obtained from the CSTR VCTK Corpus, available here: http://dx.doi.org/10.7488/ds/1994. The speech-shaped and babble noise files that were used to create this dataset are available here: http://homepages.inf.ed.ac.uk/cvbotinh/se/noises/.Valentini-Botinhao, Cassia. (2017). Noisy speech database for training speech enhancement algorithms and TTS models, 2016 [sound]. University of Edinburgh. School of Informatics. Centre for Speech Technology Research (CSTR). http://dx.doi.org/10.7488/ds/2117
Alba speech corpus
No full textSingle speaker read speech corpus of a Scottish accented female native English speaker (Alba). The corpus was recorded in four speaking styles: plain (normal read speech, around 4 hours of recordings), fast (speaking as fast as possible, around 20 mins), clear_c (computer directed speech, around 28 mins) and clear_h (hearing impaired directed speech, around 20mins). Audio files are sampled at 48kHz, segmented into sentences and saved in uncompressed formant (.wav). The underlying sentences are stored in separate text files (.txt).Valentini-Botinhao, Cassia; Yamagishi, Junichi. (2019). Alba speech corpus, [dataset]. University of Edinburgh. https://doi.org/10.7488/ds/2506
Revisiting Speech Content Privacy
Get PDFIn this paper, we discuss an important aspect of speech privacy: protecting spoken content. New capabilities from the field of machine learning provide a unique and timely opportunity to revisit speech content protection. There are many different applications of content privacy, even though this area has been under explored in speech technology research. This paper presents several scenarios that indicate a need for speech content privacy even as the specific techniques to achieve content privacy may necessarily vary. Our discussion includes several different types of content privacy including recoverable and non-recoverable content. Finally, we introduce evaluation strategies as well as describe some of the difficulties that may be encountered.<br/
Listening-test materials for "Modern speech synthesis for phonetic sciences: a discussion and an evaluation"
No full textThis data release contains listening-test materials associated with the paper "Modern speech synthesis for phonetic sciences: a discussion and an evaluation", presented at ICPhS 2019 in Melbourne, Australia.Malisz, Zofia; Henter, Gustav Eje; Valentini-Botinhao, Cassia; Watts, Oliver; Beskow, Jonas; Gustafson, Joakim. (2019). Listening-test materials for "Modern speech synthesis for phonetic sciences: a discussion and an evaluation", [dataset]. University of Edinburgh. School of Informatics. Centre for Speech Technology Research (CSTR). https://doi.org/10.7488/ds/2520
- …
