The CMU Arctic speech databases

By J. Kominek, Alan W. Black. 2004

781 citations recorded by the source corpus.

How researchers see this paper

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

1. Xu Tan
2. Wei-Ning Hsu
3. Junichi Yamagishi
4. Tao Qin
5. Jinyu Li
6. Berrak Sisman
7. Adam Polyak
8. Yu Zhang

Citations in the corpus

• SpeechT5: Unified-Modal Encoder-Decoder Pre-Training for Spoken Language Processing (2021, methodology): “We consider the many-to-many setting for the CMU Arctic (Kominek and Black, 2004), which contains speech recordings of four speakers, such as clb (female), bdl (male), slt (female), and rms (male), who read the same 1,132 phonetically balanced English utterances.”
• The Emotional Voices Database: Towards Controlling the Emotion Dimension in Voice Generation Systems (2018, methodology): “The CMU Arctic Speech Database [15] and the SIWIS French Speech Synthesis Database [13] are collections of read utterances of phonetically balanced sentences in English and French respectively.”
• Textless Speech Emotion Conversion using Decomposed and Discrete Representations (2021, background): “EmoV consists of 7000 speech utterances based on transcripts from the CMU Arctic Database (Kominek and Black, 2004).”
• Multi-Modal Data Augmentation for End-to-end ASR (2018, methodology): “For Phonestream synthetic input we constructed phone sequences using CMUDICT to which 46k words from the WSJ corpus are added[28] as the lexicon as described in section 5.”
• Cross-Lingual, Multi-Speaker Text-To-Speech Synthesis Using Neural Speaker Embedding (2019, methodology): “We also use the CMU Arctic dataset [23] which includes 7 speakers with different accents and an internal Chinese dataset includes 7 speakers with high quality speech of Mandarine for validation and testing.”
• SpeechNet: A Universal Modularized Model for Speech Processing Tasks (2021, background): “For VC, CMU Arctic (Kominek and Black, 2004) is a corpus which consists of speech utterances by 18 speakers recorded under studio conditions.”
• Epoch Extraction From Speech Signals (2008, methodology): “In the present work, the evidence for epoch locations is obtained by convolving the HE with a differenced Gaussian pulse where defines the spatial spread of the Gaussian, and is the length of the filter.”
• Many-to-Many Voice Transformer Network (2020, methodology): “We used the CMU Arctic database [69], which consists of recordings of four speakers, clb (female), bdl (male), slt (female), and rms (male), reading the same 1132 phonetically”
• An Overview of Voice Conversion and Its Challenges: From Statistical Modeling to Deep Learning (2020, methodology): “In addition to the voice conversion challenge databases described above, the CMU-Arctic database [303] and the VCTK”
• Introducing AmuS: The Amused Speech Database (2017, methodology): “For English, a subset of the CMU Arctic Speech Database [26] was used.”
• How Far Are We from Robust Voice Conversion: A Survey (2020, methodology): “Datasets We used CSTR VCTK Corpus [16], LibriTTS Corpus [24], LibriSpeech ASR Corpus [25] (LibriSpeech), CMU-ARCTIC databases [26] (CMU) and THCHS-30 [27] to evaluate performance of VC models.”
• Phonetic posteriorgrams for many-to-one voice conversion without parallel data training (2016, methodology): “The data we use for VC is the CMU ARCTIC corpus [16].”
• Automatic discovery of a phonetic inventory for unwritten languages for statistical speech synthesis (2014, methodology): “To do this, we used the EHMM labeling tool as described in [16] to do a forced alignment between phone labels and articulatory features in the RMS voice of the CMU Arctic database[17].”
• A Deep Learning Approach to Data-driven Parameterizations for Statistical Parametric Speech Synthesis (2014, methodology): “We ran experiments on 3 different voices: the RMS, and SLT voices from the CMU Arctic databases[25] and the Hindi corpus released as part of the 2014 Blizzard challenge[26].”
• Speech Enhancement for Noise-Robust Speech Synthesis Using Wasserstein GAN (2019, methodology): “As a first demonstration, we plot in Figure 1 the spectrograms from clean, noisy, SEGAN-enhanced-, and WGAN-enhanced speech using one sentence taken from the Voice bank corpus [17].”
• Towards Robust Neural Vocoding for Speech Generation: A Survey (2019, methodology): “ged in 3 days. 3. Datasets and Evaluation Metrics 3.1. Datasets for experiments The following datasets are used in our training: CMU US BDL Arctic Dataset (cmu ma), CMU US SLT Arctic Dataset (cmu fe) [22], Internal Mandarin Dataset (man fe), LibriTTS (libri) [23], and Bible (bible). CMU US BDL/SLT A”
• Espnet-TTS: Unified, Reproducible, and Integratable Open Source End-to-End Text-to-Speech Toolkit (2019, background): “05 1 6 Single Char ARCTIC [43] En 16 7 7 Adaptation Char CSMSC [44] Zn 48→ 24 1 12 Single Pinyin JNAS [45] Jp 16 306 60 Multi Phn JSUT [46] Jp 48→ 24 1 10 Single Phn JVS [47] Jp 24 100 30 Adaptation Phn LibriTTS [48] En 24 2,456 585 Multi Char LJSpeech [49] En 22.”