Citation: DOI 10.1007/s10579-011-9145-0

Collection of Audio Recordings by the Department of Computer Science at the University of Toronto from speakers with and without Dysarthtria. Useful for tasks like Audio Classification, Disease Detection, Speech Processing, etc.

Directory Structure:

F_Con : Audio Samples of female speakers from the control group, i.e., female speakers without dysarthria. 'FC01' in the folder names and the filenames refers to the first speaker, 'FC02' refers to the second speaker and so on. 'S01' refers to the first recording session with that speaker, 'S02' refers to the second session and so on. 'arrayMic' suggests that the audio was recorded with an array microphone, whereas 'headMic' suggests that the audio was recorded by a headpiece microphone.

F_Dys : Audio Samples of female speakers with dysarthria. 'F01' in the folder names and the filenames refers to the first speaker, 'F03' refers to the second speaker and so on. 'S01' refers to the first recording session with that speaker, 'S02' refers to the second session and so on. 'arrayMic' suggests that the audio was recorded with an array microphone, whereas 'headMic' suggests that the audio was recorded by a headpiece microphone.

M_Con : Audio Samples of male speakers from the control group, i.e., male speakers without dysarthria. 'MC01' in the folder names and the filenames refers to the first speaker, 'MC02' refers to the second speaker and so on. 'S01' refers to the first recording session with that speaker, 'S02' refers to the second session and so on. 'arrayMic' suggests that the audio was recorded with an array microphone, whereas 'headMic' suggests that the audio was recorded by a headpiece microphone.

M_Dys : Audio Samples of male speakers with dysarthria. 'M01' in the folder names and the filenames refers to the first speaker, 'M03' refers to the second speaker and so on. 'S01' refers to the first recording session with that speaker, 'S02' refers to the second session and so on. 'arrayMic' suggests that the audio was recorded with an array microphone, whereas 'headMic' suggests that the audio was recorded by a headpiece microphone.

Artur 1.0 is a speech database designed for the needs of automatic speech recognition for the Slovenian language. The database includes 1,067 hours of speech. 884 hours are transcribed, while the remaining 183 hours are recordings only. This repository entry includes audio files only, the transcriptions are available on http://hdl.handle.net/11356/1772.

The data are structured as follows: (1) Artur-B, read speech, 573 hours in total. It includes: (1a) Artur-B-Brani, 485 hours: Readings of sentences which were pre-selected from a 10% increment in the Gigafida 2.0 corpus. The sentences were chosen in such a way that they reflect the natural or the actual distribution of triphones in the words. They were distributed between 1,000 speakers, so that we recorded approx. 30 min in read form from each speaker. The speakers were balanced according to gender, age, region, and a small proportion of speakers were non-native speakers of Slovene. Each sentence is its own audio file and has a corresponding transcription file. (1b) Artur-B-Crkovani, 10 hours: Spellings. Speakers were asked to spell abbreviations and personal names and surnames, all chosen so that all Slovene letters were covered, plus the most common foreign letters. (1c) Artur-B-Studio, 51 hours: Designed for the development of speech synthesis. The sentences were read in a studio by a single speaker. Each sentence is its own audio file and has a corresponding transcription file. (1d) Artur-B-Izloceno, 27 hours: The recordings include different types of errors, typically, incorrect reading of sentences or a noisy environment.

(2) Artur-J, public speech, 62 hours in total. It includes: (2a) Artur-J-Splosni, 62 hours: media recordings, online recordings of conferences, workshops, education videos, etc.

(3) Artur-N, private speech, 74 hours in total. It includes: (3a) Artur-N-Obrazi, 6 hours: Speakers were asked to describe faces on pictures. Designed for a face-description domain-specific speech recognition. (3b) Artur-N-PDom, 7 hours: Speakers were asked to read pre-written sentences, as well as to express instructions for a potential smart-home system freely. Designed for a smart-home domain-specific speech recognition. (3c) Artur-N-Prosti, 61 hours: Monologues and dialogues between two persons, recorded for the purposes of the Artur database creation. Speakers were asked to conversate or explain freely on casual topics.

(4) Artur-P, parliamentary speech, 201 hours in total. It includes: (4a) Artur-P-SejeDZ, 201 hours: Speech from the Slovene National Assembly.

Further information on the database are available in the Artur-DOC file, which is part of this repository entry.

Our project, “Indonesian Media Audio Database,” is designed to establish a rich and diverse dataset tailored for training advanced machine learning models in language processing, speech recognition, and cultural analysis.

Part of the dissertation Pitch of Voiced Speech in the Short-Time Fourier Transform: Algorithms, Ground Truths, and Evaluation Methods.
© 2020, Bastian Bechtold. All rights reserved.

This dataset contains common speech and noise corpora for evaluating fundamental frequency estimation algorithms as convenient JBOF dataframes. Each corpus is available freely on its own, and allows redistribution:

• CMU-ARCTIC (BSD license) [1]
• FDA (free to download) [2]
• KEELE (free for noncommercial use) [3]
• MOCHA-TIMIT (free for noncommercial use) [4]
• PTDB-TUG (ODBL license) [5]
• NOISEX (free to download) [7]
• QUT-NOISE (CC-BY-SA license) [8]

Additionally, this dataset contains PDAs-0.0.1-py3-none-any.whl, a Python ≥ 3.6 module for Linux, containing several well-known fundamental frequency estimation algorithms:

• AUTOC [9]
• AMDF [10]
• BANA [11]
• CEP [12]
• CREPE [13]
• DIO [14]
• DNN [15]
• KALDI [16]
• MAPS
• MBSC [17]
• NLS [18]
• PEFAC [19]
• PRAAT [20]
• RAPT [21]
• SACC [22]
• SAFE [23]
• SHR [24]
• SIFT [25]
• SRH [26]
• STRAIGHT [27]
• SWIPE [28]
• YAAPT [29]
• YIN [30]

The algorithms are included in their native programming language (Matlab for BANA, DNN, MBSC, NLS, NLS2, PEFAC, RAPT, RNN, SACC, SHR, SRH, STRAIGHT, SWIPE, YAAPT, and YIN; C for KALDI, PRAAT, and SAFE; Python for AMDF, AUTOC, CEP, CREPE, MAPS, and SIFT), and adapted to a common Python interface. AMDF, AUTOC, CEP, and SIFT are our partial re-implementations as no original source code could be found.

All algorithms have been released as open source software, and are covered by their respective licenses.

All of these files are published as part of my dissertation, "Pitch of Voiced Speech in the Short-Time Fourier Transform: Algorithms, Ground Truths, and Evaluation Methods", and in support of the Replication Dataset for Fundamental Frequency Estimation.

References:

1. John Kominek and Alan W Black. CMU ARCTIC database for speech synthesis, 2003.
2. Paul C Bagshaw, Steven Hiller, and Mervyn A Jack. Enhanced Pitch Tracking and the Processing of F0 Contours for Computer Aided Intonation Teaching. In EUROSPEECH, 1993.
3. F Plante, Georg F Meyer, and William A Ainsworth. A Pitch Extraction Reference Database. In Fourth European Conference on Speech Communication and Technology, pages 837–840, Madrid, Spain, 1995.
4. Alan Wrench. MOCHA MultiCHannel Articulatory database: English, November 1999.
5. Gregor Pirker, Michael Wohlmayr, Stefan Petrik, and Franz Pernkopf. A Pitch Tracking Corpus with Evaluation on Multipitch Tracking Scenario. page 4, 2011.
6. John S. Garofolo, Lori F. Lamel, William M. Fisher, Jonathan G. Fiscus, David S. Pallett, Nancy L. Dahlgren, and Victor Zue. TIMIT Acoustic-Phonetic Continuous Speech Corpus, 1993.
7. Andrew Varga and Herman J.M. Steeneken. Assessment for automatic speech recognition: II. NOISEX-92: A database and an experiment to study the effect of additive noise on speech recog- nition systems. Speech Communication, 12(3):247–251, July 1993.
8. David B. Dean, Sridha Sridharan, Robert J. Vogt, and Michael W. Mason. The QUT-NOISE-TIMIT corpus for the evaluation of voice activity detection algorithms. Proceedings of Interspeech 2010, 2010.
9. Man Mohan Sondhi. New methods of pitch extraction. Audio and Electroacoustics, IEEE Transactions on, 16(2):262—266, 1968.
10. Myron J. Ross, Harry L. Shaffer, Asaf Cohen, Richard Freudberg, and Harold J. Manley. Average magnitude difference function pitch extractor. Acoustics, Speech and Signal Processing, IEEE Transactions on, 22(5):353—362, 1974.
11. Na Yang, He Ba, Weiyang Cai, Ilker Demirkol, and Wendi Heinzelman. BaNa: A Noise Resilient Fundamental Frequency Detection Algorithm for Speech and Music. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(12):1833–1848, December 2014.
12. Michael Noll. Cepstrum Pitch Determination. The Journal of the Acoustical Society of America, 41(2):293–309, 1967.
13. Jong Wook Kim, Justin Salamon, Peter Li, and Juan Pablo Bello. CREPE: A Convolutional Representation for Pitch Estimation. arXiv:1802.06182 [cs, eess, stat], February 2018. arXiv: 1802.06182.
14. Masanori Morise, Fumiya Yokomori, and Kenji Ozawa. WORLD: A Vocoder-Based High-Quality Speech Synthesis System for Real-Time Applications. IEICE Transactions on Information and Systems, E99.D(7):1877–1884, 2016.
15. Kun Han and DeLiang Wang. Neural Network Based Pitch Tracking in Very Noisy Speech. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(12):2158–2168, Decem- ber 2014.
16. Pegah Ghahremani, Bagher BabaAli, Daniel Povey, Korbinian Riedhammer, Jan Trmal, and Sanjeev Khudanpur. A pitch extraction algorithm tuned for automatic speech recognition. In Acoustics, Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on, pages 2494–2498. IEEE, 2014.
17. Lee Ngee Tan and Abeer Alwan. Multi-band summary correlogram-based pitch detection for noisy speech. Speech Communication, 55(7-8):841–856, September 2013.
18. Jesper Kjær Nielsen, Tobias Lindstrøm Jensen, Jesper Rindom Jensen, Mads Græsbøll Christensen, and Søren Holdt Jensen. Fast fundamental frequency estimation: Making a statistically efficient estimator computationally efficient. Signal Processing, 135:188–197, June 2017.
19. Sira Gonzalez and Mike Brookes. PEFAC - A Pitch Estimation Algorithm Robust to High Levels of Noise. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(2):518—530, February 2014.
20. Paul Boersma. Accurate short-term analysis of the fundamental frequency and the harmonics-to-noise ratio of a sampled sound. In Proceedings of the institute of phonetic sciences, volume 17, page 97—110. Amsterdam, 1993.
21. David Talkin. A robust algorithm for pitch tracking (RAPT). Speech coding and synthesis, 495:518, 1995.
22. Byung Suk Lee and Daniel PW Ellis. Noise robust pitch tracking by subband autocorrelation classification. In Interspeech, pages 707–710, 2012.
23. Wei Chu and Abeer Alwan. SAFE: a statistical algorithm for F0 estimation for both clean and noisy speech. In INTERSPEECH, pages 2590–2593, 2010.
24. Xuejing Sun. Pitch determination and voice quality analysis using subharmonic-to-harmonic ratio. In Acoustics, Speech, and Signal Processing (ICASSP), 2002 IEEE International Conference on, volume 1, page I—333. IEEE, 2002.
25. Markel. The SIFT algorithm for fundamental frequency estimation. IEEE Transactions on Audio and Electroacoustics, 20(5):367—377, December 1972.
26. Thomas Drugman and Abeer Alwan. Joint Robust Voicing Detection and Pitch Estimation Based on Residual Harmonics. In Interspeech, page 1973—1976, 2011.
27. Hideki Kawahara, Masanori Morise, Toru Takahashi, Ryuichi Nisimura, Toshio Irino, and Hideki Banno. TANDEM-STRAIGHT: A temporally stable power spectral representation for periodic signals and applications to interference-free spectrum, F0, and aperiodicity estimation. In Acous- tics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on, pages 3933–3936. IEEE, 2008.
28. Arturo Camacho. SWIPE: A sawtooth waveform inspired pitch estimator for speech and music. PhD thesis, University of Florida, 2007.
29. Kavita Kasi and Stephen A. Zahorian. Yet Another Algorithm for Pitch Tracking. In IEEE International Conference on Acoustics Speech and Signal Processing, pages I–361–I–364, Orlando, FL, USA, May 2002. IEEE.
30. Alain de Cheveigné and Hideki Kawahara. YIN, a fundamental frequency estimator for speech and music. The Journal

== Quick facts ==

The most up-to-date and comprehensive podcast database available All languages & All countries Includes over 3,500,000 podcasts Features 35+ data fields , such as basic metadata, global rank, RSS feed (with audio URLs), Spotify links, and more Delivered in SQLite format Learn how we build a high quality podcast database: https://www.listennotes.help/article/105-high-quality-podcast-database-from-listen-notes

== Use Cases ==

AI training, including speech recognition, generative AI, voice cloning / synthesis, and news analysis Alternative data for investment research, such as sentiment analysis of executive interviews, market research and tracking investment themes PR and marketing, including social monitoring, content research, outreach, and guest booking ...

== Data Attributes ==

See the full list of data attributes on this page: https://www.listennotes.com/podcast-datasets/fields/?filter=podcast_only

How to access podcast audio files: Our dataset includes RSS feed URLs for all podcasts. You can retrieve audio for over 170 million episodes directly from these feeds. With access to the raw audio, you’ll have high-quality podcast speech data ideal for AI training and related applications.

== Custom Offers ==

We can provide custom datasets based on your needs, such as language-specific data, daily/weekly/monthly update frequency, or one-time purchases.

We also provide a RESTful API at PodcastAPI.com

Contact us: hello@listennotes.com

== Need Help? ==

If you have any questions about our products, feel free to reach out hello@listennotes.com

== About Listen Notes, Inc. ==

Since 2017, Listen Notes, Inc. has provided the leading podcast search engine and podcast database.

• Can you predict which speaker is talking?
• Can you predict what they are saying? This dataset makes all of these possible. Perfect for a school project, research project, or resume builder.

File information

• train.csv - file containing all the data you need for training, with 4 columns, id (file id), file_path(path to .wav files), speech(transcription of audio file), and speaker (target column)
• test.csv - file containing all the data you need to test your model (20% of total audio files), it has the same columns as train.csv
• train/ - Folder with training data, subdivided with Speaker's folders
  • aew/ - Folder containing audio files in .wav format for speaker aew
  • ...
• test/ - Folder containing audio files for test data.

Column description

More Details

The CMU_ARCTIC databases were constructed at the Language Technologies Institute at Carnegie Mellon University as phonetically balanced, US-English single-speaker databases designed for unit selection speech synthesis research. A detailed report on the structure and content of the database and the recording environment etc is available as a Carnegie Mellon University, Language Technologies Institute Tech Report CMU-LTI-03-177 and is also available here.

The databases consist of around 1150 utterances carefully selected from out-of-copyright texts from Project Gutenberg. The databases include US English male (bdl) and female (slt) speakers (both experienced voice talent) as well as other accented speakers.

The 1132 sentence prompt list is available from cmuarctic.data

The distributions include 16KHz waveform and simultaneous EGG signals. Full phonetically labeling was performed by the CMU Sphinx using the FestVox based labeling scripts. Complete runnable Festival Voices are included with the database distributions, as examples though better voices can be made by improving labeling, etc.

Acknowledgements

This work was partially supported by the U.S. National Science Foundation under Grant No. 0219687, "ITR/CIS Evaluation and Personalization of Synthetic Voices". Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

The Aurora project was originally set up to establish a worldwide standard for the feature extraction software which forms the core of the front-end of a DSR (Distributed Speech Recognition) system. The AURORA-5 database has been mainly developed to investigate the influence on the performance of automatic speech recognition for a hands-free speech input in noisy room environments. Furthermore two test conditions are included to study the influence of transmitting the speech in a mobile communication system.The earlier three Aurora experiments had a focus on additive noise and the influence of some telephone frequency characteristics. Aurora-5 tries to cover all effects as they occur in realistic application scenarios. The focus was put on two scenarios. The first one is the hands-free speech input in the noisy car environment with the intention of controlling either devices in the car itself or retrieving information from a remote speech server over the telephone. The second one covers the hands-free speech input in a type of office or in a type of living room to control e.g. a telephone device or some audio/video equipment.The AURORA-5 database contains the following data:•Artificially distorted versions of the recordings from adult speakers in the TI-Digits speech database downsampled at a sampling frequency of 8000 Hz. The distortions consist of: - additive background noise, - the simulation of a hands-free speech input in rooms, - the simulation of transmitting speech over cellular telephone networks.•A subset of recordings from the meeting recorder project at the International Computer Science Institute. The recordings contain sequences of digits uttered by different speakers in hands-free mode in a meeting room.•A set of scripts for running recognition experiments on the above mentioned speech data. The experiments are based on the usage of the freely available software package HTK where HTK is not part of this resource.Further information is also available at the following address: http://aurora.hsnr.de

Description

The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) contains 7356 files (total size: 24.8 GB). The dataset contains 24 professional actors (12 female, 12 male), vocalizing two lexically-matched statements in a neutral North American accent. Speech includes calm, happy, sad, angry, fearful, surprise, and disgust expressions, and song contains calm, happy, sad, angry, and fearful emotions. Each expression is produced at two levels of emotional intensity (normal, strong), with an additional neutral expression. All conditions are available in three modality formats: Audio-only (16bit, 48kHz .wav), Audio-Video (720p H.264, AAC 48kHz, .mp4), and Video-only (no sound). Note, there are no song files for Actor_18.

The RAVDESS was developed by Dr Steven R. Livingstone, who now leads the Affective Data Science Lab, and Dr Frank A. Russo who leads the SMART Lab.

Citing the RAVDESS

The RAVDESS is released under a Creative Commons Attribution license, so please cite the RAVDESS if it is used in your work in any form. Published academic papers should use the academic paper citation for our PLoS1 paper. Personal works, such as machine learning projects/blog posts, should provide a URL to this Zenodo page, though a reference to our PLoS1 paper would also be appreciated.

Academic paper citation

Livingstone SR, Russo FA (2018) The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS): A dynamic, multimodal set of facial and vocal expressions in North American English. PLoS ONE 13(5): e0196391. https://doi.org/10.1371/journal.pone.0196391.

Personal use citation

Include a link to this Zenodo page - https://zenodo.org/record/1188976

Commercial Licenses

Commercial licenses for the RAVDESS can be purchased. For more information, please visit our license page of fees, or contact us at ravdess@gmail.com.

Contact Information

If you would like further information about the RAVDESS, to purchase a commercial license, or if you experience any issues downloading files, please contact us at ravdess@gmail.com.

Example Videos

Watch a sample of the RAVDESS speech and song videos.

Emotion Classification Users

If you're interested in using machine learning to classify emotional expressions with the RAVDESS, please see our new RAVDESS Facial Landmark Tracking data set [Zenodo project page].

Construction and Validation

Full details on the construction and perceptual validation of the RAVDESS are described in our PLoS ONE paper - https://doi.org/10.1371/journal.pone.0196391.

The RAVDESS contains 7356 files. Each file was rated 10 times on emotional validity, intensity, and genuineness. Ratings were provided by 247 individuals who were characteristic of untrained adult research participants from North America. A further set of 72 participants provided test-retest data. High levels of emotional validity, interrater reliability, and test-retest intrarater reliability were reported. Validation data is open-access, and can be downloaded along with our paper from PLoS ONE.

Contents

Audio-only files

Audio-only files of all actors (01-24) are available as two separate zip files (~200 MB each):

• Speech file (Audio_Speech_Actors_01-24.zip, 215 MB) contains 1440 files: 60 trials per actor x 24 actors = 1440.
• Song file (Audio_Song_Actors_01-24.zip, 198 MB) contains 1012 files: 44 trials per actor x 23 actors = 1012.

Audio-Visual and Video-only files

Video files are provided as separate zip downloads for each actor (01-24, ~500 MB each), and are split into separate speech and song downloads:

• Speech files (Video_Speech_Actor_01.zip to Video_Speech_Actor_24.zip) collectively contains 2880 files: 60 trials per actor x 2 modalities (AV, VO) x 24 actors = 2880.
• Song files (Video_Song_Actor_01.zip to Video_Song_Actor_24.zip) collectively contains 2024 files: 44 trials per actor x 2 modalities (AV, VO) x 23 actors = 2024.

File Summary

In total, the RAVDESS collection includes 7356 files (2880+2024+1440+1012 files).

File naming convention

Each of the 7356 RAVDESS files has a unique filename. The filename consists of a 7-part numerical identifier (e.g., 02-01-06-01-02-01-12.mp4). These identifiers define the stimulus characteristics:

Filename identifiers

• Modality (01 = full-AV, 02 = video-only, 03 = audio-only).
• Vocal channel (01 = speech, 02 = song).
• Emotion (01 = neutral, 02 = calm, 03 = happy, 04 = sad, 05 = angry, 06 = fearful, 07 = disgust, 08 = surprised).
• Emotional intensity (01 = normal, 02 = strong). NOTE: There is no strong intensity for the 'neutral' emotion.
• Statement (01 = "Kids are talking by the door", 02 = "Dogs are sitting by the door").
• Repetition (01 = 1st repetition, 02 = 2nd repetition).
• Actor (01 to 24. Odd numbered actors are male, even numbered actors are female).

Filename example: 02-01-06-01-02-01-12.mp4

1. Video-only (02)
2. Speech (01)
3. Fearful (06)
4. Normal intensity (01)
5. Statement "dogs" (02)
6. 1st Repetition (01)
7. 12th Actor (12)
8. Female, as the actor ID number is even.

License information

The RAVDESS is released under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License, CC BY-NC-SA 4.0

Commercial licenses for the RAVDESS can also be purchased. For more information, please visit our license fee page, or contact us at ravdess@gmail.com.

Related Data sets

• RAVDESS Facial Landmark Tracking data set [Zenodo project page].

Ultrasound has been used to render animations of articulation during speech production, specifically tongue movement, for visual feedback used in intervention for articulation disorders and speech recognition. Nevertheless, the availability of high-quality audio-ultrasound datasets remains scarce. The present study, therefore, aims to construct a multimodal database designed for Mandarin speech. The dataset integrates synchronized ultrasound images of lingual movement, and the corresponding audio recordings and text annotations elicited from 43 healthy speakers and 11 patients with dysarthria through speech tasks (including monophthong vowels, monosyllables, and sentences), with a total duration of 22.31 hours. During production, a high-resolution (920×700 at 60 fps) ultrasound device and a high-fidelity microphone were used to simultaneously record tongue motion and audio signals, maintaining sync via experimental setup. In addition, a customized helmet structure was employed to stabilize the ultrasound probe, precisely controlling for head movement, minimizing displacement interference, and ensuring spatial stability of the images. The proposed database carries apparent values in automatic speech recognition, silent interface development, and research in speech pathology and linguistics.

== Quick facts ==

The most up-to-date and comprehensive podcast database available Includes over 3,500,000 podcasts and over 176 million episodes (including direct playable audio urls) Features 35+ data fields , such as basic metadata, global rank, RSS feed (with audio URLs), Spotify links, and more Delivered in SQLite format

== Use Cases ==

AI training, including speech recognition, generative AI, voice cloning / synthesis, and news analysis Alternative data for investment research, such as sentiment analysis of executive interviews, market research and tracking investment themes PR and marketing, including social monitoring, content research, outreach, and guest booking ...

== Custom Offers ==

We can provide custom datasets based on your needs, such as language-specific data, daily/weekly/monthly update frequency, or one-time purchases.

We also provide a RESTful API at PodcastAPI.com

Contact us: hello@listennotes.com

== Need Help? ==

If you have any questions about our products, feel free to reach out hello@listennotes.com

== About Listen Notes, Inc. ==

Since 2017, Listen Notes, Inc. has provided the leading podcast search engine and podcast database.

Part of the dissertation Pitch of Voiced Speech in the Short-Time Fourier Transform: Algorithms, Ground Truths, and Evaluation Methods.
© 2020, Bastian Bechtold. All rights reserved.

Estimating the fundamental frequency of speech remains an active area of research, with varied applications in speech recognition, speaker identification, and speech compression. A vast number of algorithms for estimatimating this quantity have been proposed over the years, and a number of speech and noise corpora have been developed for evaluating their performance. The present dataset contains estimated fundamental frequency tracks of 25 algorithms, six speech corpora, two noise corpora, at nine signal-to-noise ratios between -20 and 20 dB SNR, as well as an additional evaluation of synthetic harmonic tone complexes in white noise.

The dataset also contains pre-calculated performance measures both novel and traditional, in reference to each speech corpus’ ground truth, the algorithms’ own clean-speech estimate, and our own consensus truth. It can thus serve as the basis for a comparison study, or to replicate existing studies from a larger dataset, or as a reference for developing new fundamental frequency estimation algorithms. All source code and data is available to download, and entirely reproducible, albeit requiring about one year of processor-time.

Included Code and Data

• ground truth data.zip is a JBOF dataset of fundamental frequency estimates and ground truths of all speech files in the following corpora:
  • CMU-ARCTIC (consensus truth) [1]
  • FDA (corpus truth and consensus truth) [2]
  • KEELE (corpus truth and consensus truth) [3]
  • MOCHA-TIMIT (consensus truth) [4]
  • PTDB-TUG (corpus truth and consensus truth) [5]
  • TIMIT (consensus truth) [6]
• noisy speech data.zip is a JBOF datasets of fundamental frequency estimates of speech files mixed with noise from the following corpora:
  • NOISEX [7]
  • QUT-NOISE [8]
• synthetic speech data.zip is a JBOF dataset of fundamental frequency estimates of synthetic harmonic tone complexes in white noise.
• noisy_speech.pkl and synthetic_speech.pkl are pickled Pandas dataframes of performance metrics derived from the above data for the following list of fundamental frequency estimation algorithms:
  • AUTOC [9]
  • AMDF [10]
  • BANA [11]
  • CEP [12]
  • CREPE [13]
  • DIO [14]
  • DNN [15]
  • KALDI [16]
  • MAPS
  • MBSC [17]
  • NLS [18]
  • PEFAC [19]
  • PRAAT [20]
  • RAPT [21]
  • SACC [22]
  • SAFE [23]
  • SHR [24]
  • SIFT [25]
  • SRH [26]
  • STRAIGHT [27]
  • SWIPE [28]
  • YAAPT [29]
  • YIN [30]
• noisy speech evaluation.py and synthetic speech evaluation.py are Python programs to calculate the above Pandas dataframes from the above JBOF datasets. They calculate the following performance measures:
  • Gross Pitch Error (GPE), the percentage of pitches where the estimated pitch deviates from the true pitch by more than 20%.
  • Fine Pitch Error (FPE), the mean error of grossly correct estimates.
  • High/Low Octave Pitch Error (OPE), the percentage pitches that are GPEs and happens to be at an integer multiple of the true pitch.
  • Gross Remaining Error (GRE), the percentage of pitches that are GPEs but not OPEs.
  • Fine Remaining Bias (FRB), the median error of GREs.
  • True Positive Rate (TPR), the percentage of true positive voicing estimates.
  • False Positive Rate (FPR), the percentage of false positive voicing estimates.
  • False Negative Rate (FNR), the percentage of false negative voicing estimates.
  • F₁, the harmonic mean of precision and recall of the voicing decision.
• Pipfile is a pipenv-compatible pipfile for installing all prerequisites necessary for running the above Python programs.

The Python programs take about an hour to compute on a fast 2019 computer, and require at least 32 Gb of memory.

References:

1. John Kominek and Alan W Black. CMU ARCTIC database for speech synthesis, 2003.
2. Paul C Bagshaw, Steven Hiller, and Mervyn A Jack. Enhanced Pitch Tracking and the Processing of F0 Contours for Computer Aided Intonation Teaching. In EUROSPEECH, 1993.
3. F Plante, Georg F Meyer, and William A Ainsworth. A Pitch Extraction Reference Database. In Fourth European Conference on Speech Communication and Technology, pages 837–840, Madrid, Spain, 1995.
4. Alan Wrench. MOCHA MultiCHannel Articulatory database: English, November 1999.
5. Gregor Pirker, Michael Wohlmayr, Stefan Petrik, and Franz Pernkopf. A Pitch Tracking Corpus with Evaluation on Multipitch Tracking Scenario. page 4, 2011.
6. John S. Garofolo, Lori F. Lamel, William M. Fisher, Jonathan G. Fiscus, David S. Pallett, Nancy L. Dahlgren, and Victor Zue. TIMIT Acoustic-Phonetic Continuous Speech Corpus, 1993.
7. Andrew Varga and Herman J.M. Steeneken. Assessment for automatic speech recognition: II. NOISEX-92: A database and an experiment to study the effect of additive noise on speech recog- nition systems. Speech Communication, 12(3):247–251, July 1993.
8. David B. Dean, Sridha Sridharan, Robert J. Vogt, and Michael W. Mason. The QUT-NOISE-TIMIT corpus for the evaluation of voice activity detection algorithms. Proceedings of Interspeech 2010, 2010.
9. Man Mohan Sondhi. New methods of pitch extraction. Audio and Electroacoustics, IEEE Transactions on, 16(2):262—266, 1968.
10. Myron J. Ross, Harry L. Shaffer, Asaf Cohen, Richard Freudberg, and Harold J. Manley. Average magnitude difference function pitch extractor. Acoustics, Speech and Signal Processing, IEEE Transactions on, 22(5):353—362, 1974.
11. Na Yang, He Ba, Weiyang Cai, Ilker Demirkol, and Wendi Heinzelman. BaNa: A Noise Resilient Fundamental Frequency Detection Algorithm for Speech and Music. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(12):1833–1848, December 2014.
12. Michael Noll. Cepstrum Pitch Determination. The Journal of the Acoustical Society of America, 41(2):293–309, 1967.
13. Jong Wook Kim, Justin Salamon, Peter Li, and Juan Pablo Bello. CREPE: A Convolutional Representation for Pitch Estimation. arXiv:1802.06182 [cs, eess, stat], February 2018. arXiv: 1802.06182.
14. Masanori Morise, Fumiya Yokomori, and Kenji Ozawa. WORLD: A Vocoder-Based High-Quality Speech Synthesis System for Real-Time Applications. IEICE Transactions on Information and Systems, E99.D(7):1877–1884, 2016.
15. Kun Han and DeLiang Wang. Neural Network Based Pitch Tracking in Very Noisy Speech. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 22(12):2158–2168, Decem- ber 2014.
16. Pegah Ghahremani, Bagher BabaAli, Daniel Povey, Korbinian Riedhammer, Jan Trmal, and Sanjeev Khudanpur. A pitch extraction algorithm tuned for automatic speech recognition. In Acoustics, Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on, pages 2494–2498. IEEE, 2014.
17. Lee Ngee Tan and Abeer Alwan. Multi-band summary correlogram-based pitch detection for noisy speech. Speech Communication, 55(7-8):841–856, September 2013.
18. Jesper Kjær Nielsen, Tobias Lindstrøm Jensen, Jesper Rindom Jensen, Mads Græsbøll Christensen, and Søren Holdt Jensen. Fast fundamental frequency estimation: Making a statistically

The corpus contains 26 pairs of Afghanistan Southern Pashto spontaneous conversational speech, which were from 52 speakers (27 males and 25 females). For this collection, 2 speakers of each group performed the recording in separate quiet rooms. 21 topics were contained in this database. The audio duration is 160.3 hours and the speech hour is about 50.8 hours, including the reasonable leading and trailing silence. The total size of this database is 8.6 GB.

Our Laboratory of Artificial Neural Network Applications (LANNA) in the Czech Technical University in Prague (head of the laboratory is professor Jana Tučková) collaborates on a project with the Department of Paediatric Neurology, 2nd Faculty of Medicine of Charles University in Prague and with the Motol University Hospital (head of clinic is professor Vladimír Komárek), which focuses on the study of children with SLI.The speech database contains two subgroups of recordings of children's speech from different types of speakers. The first subgroup (healthy) consists of recordings of children without speech disorders; the second subgroup (patients) consists of recordings of children with SLI. These children have different degrees of severity (1 – mild, 2 – moderate, and 3 – severe). The speech therapists and specialists from Motol Hospital decided upon this classification. The children’s speech was recorded in the period 2003-2013. These databases were commonly created in a schoolroom or a speech therapist’s consulting room, in the presence of surrounding background noise. This situation simulates the natural environment in which the children live, and is important for capturing the normal behavior of children. The database of healthy children’s speech was created as a referential database for the computer processing of children’s speech. It was recorded on the SONY digital Dictaphone (sampling frequency, fs = 16 kHz, 16-bit resolution in stereo mode in the standardized wav format) and on the MD SONY MZ-N710 (sampling frequency, fs = 44.1 kHz, 16-bit resolution in stereo mode in the standardized wav format). The corpus was recorded in the natural environment of a schoolroom and in a clinic. This subgroup contains a total of 44 native Czech participants (15 boys, 29 girls) aged 4 to 12 years, and was recorded during the period 2003–2005. The database of children with SLI was recorded in a private speech therapist’s office. The children’s speech is captured by means of a SHURE lapel microphone using the solution by the company AVID (MBox – USB AD/DA converter and ProTools LE software) on an Apple laptop (iBook G4). The sound recordings are saved in the standardized wav format. The sampling frequency is set to 44.1 kHz with 16-bit resolution in mono mode. This subgroup contains a total of 54 native Czech participants (35 boys, 19 girls) aged 6 to 12 years, and was recorded during the period 2009–2013. This package contains wav data sets for development and testing methods for detection children with SLI.Software pack:FORANA - was developed the original software FORANA for formants analysis. It is based on the MATLAB programming environment. The development of this software was mainly driven by the need to have the ability to complete formant analysis correctly and full automation of the process of extracting formants from the recorded speech signals. Development of this application is still running. Software was developed in the LANNA at CTU FEE in Prague.LABELING - the program LABELING is used for segmentation of the speech signal. It is a part of SOMLab program system. Software was developed in the LANNA at CTU FEE in Prague.PRAAT - is an acoustic analysis software. The Praat program was created by Paul Boersma and David Weenink of the Institute of Phonetics Sciences of the University of Amsterdam. Home page:http://www.praat.org or http://www.fon.hum.uva.nl/praat/.openSMILE - The openSMILE feature extration tool enables you to extract large audio feature spaces in realtime. It combines features from Music Information Retrieval and Speech Processing. SMILE is an acronym forSpeech & Music Interpretation by Large-space Extraction. It is written in C++ and is available as both a standalone commandline executable as well as a dynamic library. The main features of openSMILE are its capability of on-line incremental processing and its modularity. Feature extractor components can be freely interconnected to create new and custom features, all via a simple configuration file. New components can be added to openSMILE via an easy binary plugin interface and a comprehensive API. Citing: Florian Eyben, Martin Wöllmer, Björn Schuller: "openSMILE - The Munich Versatile and Fast Open-Source Audio Feature Extractor", In Proc. ACM Multimedia (MM), ACM, Florence, Italy, ACM, ISBN 978-1-60558-933-6, pp. 1459-1462, October 2010. doi:10.1145/1873951.1874246

SpeakingFaces is a publicly-available large-scale dataset developed to support multimodal machine learning research in contexts that utilize a combination of thermal, visual, and audio data streams; examples include human-computer interaction (HCI), biometric authentication, recognition systems, domain transfer, and speech recognition. SpeakingFaces is comprised of well-aligned high-resolution thermal and visual spectra image streams of fully-framed faces synchronized with audio recordings of each subject speaking approximately 100 imperative phrases.

Database description:

The written and spoken digits database is not a new database but a constructed database from existing ones, in order to provide a ready-to-use database for multimodal fusion [1].

The written digits database is the original MNIST handwritten digits database [2] with no additional processing. It consists of 70000 images (60000 for training and 10000 for test) of 28 x 28 = 784 dimensions.

The spoken digits database was extracted from Google Speech Commands [3], an audio dataset of spoken words that was proposed to train and evaluate keyword spotting systems. It consists of 105829 utterances of 35 words, amongst which 38908 utterances of the ten digits (34801 for training and 4107 for test). A pre-processing was done via the extraction of the Mel Frequency Cepstral Coefficients (MFCC) with a framing window size of 50 ms and frame shift size of 25 ms. Since the speech samples are approximately 1 s long, we end up with 39 time slots. For each one, we extract 12 MFCC coefficients with an additional energy coefficient. Thus, we have a final vector of 39 x 13 = 507 dimensions. Standardization and normalization were applied on the MFCC features.

To construct the multimodal digits dataset, we associated written and spoken digits of the same class respecting the initial partitioning in [2] and [3] for the training and test subsets. Since we have less samples for the spoken digits, we duplicated some random samples to match the number of written digits and have a multimodal digits database of 70000 samples (60000 for training and 10000 for test).

The dataset is provided in six files as described below. Therefore, if a shuffle is performed on the training or test subsets, it must be performed in unison with the same order for the written digits, spoken digits and labels.

Files:

• data_wr_train.npy: 60000 samples of 784-dimentional written digits for training;
• data_sp_train.npy: 60000 samples of 507-dimentional spoken digits for training;
• labels_train.npy: 60000 labels for the training subset;
• data_wr_test.npy: 10000 samples of 784-dimentional written digits for test;
• data_sp_test.npy: 10000 samples of 507-dimentional spoken digits for test;
• labels_test.npy: 10000 labels for the test subset.

References:

1. Khacef, L. et al. (2020), "Brain-Inspired Self-Organization with Cellular Neuromorphic Computing for Multimodal Unsupervised Learning".
2. LeCun, Y. & Cortes, C. (1998), “MNIST handwritten digit database”.
3. Warden, P. (2018), “Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition”.

The Emotional Voice Messages (EMOVOME) database is a speech dataset collected for emotion recognition in real-world conditions. It contains 999 spontaneous voice messages from 100 Spanish speakers, collected from real conversations on a messaging app. EMOVOME includes both expert and non-expert emotional annotations, covering valence and arousal dimensions, along with emotion categories for the expert annotations. Detailed participant information is provided, including sociodemographic data and personality trait assessments using the NEO-FFI questionnaire. Moreover, EMOVOME provides audio recordings of participants reading a given text, as well as transcriptions of all 999 voice messages. Additionally, baseline models for valence and arousal recognition are provided, utilizing both speech and audio transcriptions.

Description

For details on the EMOVOME database, please refer to the article:

"EMOVOME Database: Advancing Emotion Recognition in Speech Beyond Staged Scenarios". Lucía Gómez-Zaragozá, Rocío del Amor, María José Castro-Bleda, Valery Naranjo, Mariano Alcañiz Raya, Javier Marín-Morales. (pre-print available in https://doi.org/10.48550/arXiv.2403.02167)

Content

The Zenodo repository contains four files:

EMOVOME_agreement.pdf: agreement file required to access the original audio files, detailed in section Usage Notes.

labels.csv: ratings of the three non-experts and the expert annotator, independently and combined.

participants_ids.csv: table mapping each numerical file ID to its corresponding alphanumeric participant ID.

transcriptions.csv: transcriptions of each audio.

The repository also includes three folders:

Audios: it contains the file features_eGeMAPSv02.csv corresponding to the standard acoustic feature set used in the baseline model, and two folders:

Lecture: contains the audio files corresponding to the text readings, with each file named according to the participant's ID.

Emotions: contains the voice recordings from the messaging app provided by the user, named with a file ID.

Questionnaires: it contains two files: 1) sociodemographic_spanish.csv and sociodemographic_english.csv are the sociodemographic data of participants in Spanish and English, respectively, including the demographic information; and 2) NEO-FFI_spanish.csv includes the participants’ answers to the Spanish version of the NEO-FFI questionnaire. The three files include a column indicating the participant's ID to link the information.

Baseline_emotion_recognition: it includes three files and two folders. The file partitions.csv specifies the proposed data partition. Particularly, the dataset is divided into 80% for development and 20% for testing using a speaker-independent approach, i.e., samples from the same speaker are not included in both development and test. The development set includes 80 participants (40 female, 40 male) containing the following distribution of labels: 241 negative, 305 neutral and 261 positive valence; and 148 low, 328 neutral and 331 high arousal. The test set includes 20 participants (10 female, 10 male) with the distribution of labels that follows: 57 negative, 62 neutral and 73 positive valence; and 13 low, 70 neutral and 109 high arousal. Files baseline_speech.ipynb and baseline_text.ipynb contain the code used to create the baseline emotion recognition models based on speech and text, respectively. The actual trained models for valence and arousal prediction are provided in folders models_speech and models_text.

Audio files in “Lecture” and “Emotions” are only provided to the users that complete the agreement file in section Usage Notes. Audio files are in Ogg Vorbis format at 16-bit and 44.1 kHz or 48 kHz. The total size of the “Audios” folder is about 213 MB.

Usage Notes

All the data included in the EMOVOME database is publicly available under the Creative Commons Attribution 4.0 International license. The only exception is the original raw audio files, for which an additional step is required as a security measure to safeguard the speakers' privacy. To request access, interested authors should first complete and sign the agreement file EMOVOME_agreement.pdf and send it to the corresponding author (jamarmo@htech.upv.es). The data included in the EMOVOME database is expected to be used for research purposes only. Therefore, the agreement file states that the authors are not allowed to share the data with profit-making companies or organisations. They are also not expected to distribute the data to other research institutions; instead, they are suggested to kindly refer interested colleagues to the corresponding author of this article. By agreeing to the terms of the agreement, the authors also commit to refraining from publishing the audio content on the media (such as television and radio), in scientific journals (or any other publications), as well as on other platforms on the internet. The agreement must bear the signature of the legally authorised representative of the research institution (e.g., head of laboratory/department). Once the signed agreement is received and validated, the corresponding author will deliver the "Audios" folder containing the audio files through a download procedure. A direct connection between the EMOVOME authors and the applicants guarantees that updates regarding additional materials included in the database can be received by all EMOVOME users.

AbstractIntroduction The Audiovisual Database of Spoken American English, Linguistic Data Consortium (LDC) catalog number LDC2009V01 and isbn 1-58563-496-4, was developed at Butler University, Indianapolis, IN in 2007 for use by a a variety of researchers to evaluate speech production and speech recognition. It contains approximately seven hours of audiovisual recordings of fourteen American English speakers producing syllables, word lists and sentences used in both academic and clinical settings. All talkers were from the North Midland dialect region -- roughly defined as Indianapolis and north within the state of Indiana -- and had lived in that region for the majority of the time from birth to 18 years of age. Each participant read 238 different words and 166 different sentences. The sentences spoken were drawn from the following sources: Central Institute for the Deaf (CID) Everyday Sentences (Lists A-J) Northwestern University Auditory Test No. 6 (Lists I-IV) Vowels in /hVd/ context (separate words) Texas Instruments/Massachusetts Institute for Technology (TIMIT) sentences The CID Everyday Sentences were created in the 1950s from a sample developed by the Armed Forces National Research Committee on Hearing and Bio-Acoustics. They are considered to represent everyday American speech and have the following characteristics: the vocabulary is appropriate to adults; the words appear with high frequency in one or more of the well-known word counts of the English language; proper names and proper nouns are not used; common non-slang idioms and contractions are used freely; phonetic loading and "tongue-twisting" are avoided; redundancy is high; the level of abstraction is low; and grammatical structure varies freely. Northwestern University Auditory Test No. 6 is a phonemically-balanced set of monosyllabic English words used clinically to test speech perception in adults with hearing loss. The /hVd/ vowel list was created to elicit all of the vowel sounds of American English. The TIMIT sentences are a subset (34 sentences) of the 2342 phonetically-rich sentences read by speakers in the TIMIT Acoustic-Phonetic Continuous Speech Corpus LDC93S1. TIMIT was designed to provide speech data for the acquisition of acoustic-phonetic knowledge and for the development and evaluation of automatic speech recognition systems. TIMIT speakers were from eight dialect regions of the United States. The Audiovisual Database of Spoken American English will be of interest in various disciplines: to linguists for studies of phonetics, phonology, and prosody of American English; to speech scientists for investigations of motor speech production and auditory-visual speech perception; to engineers and computer scientists for investigations of machine audio-visual speech recognition (AVSR); and to speech and hearing scientists for clinical purposes, such as the examination and improvement of speech perception by listeners with hearing loss. Data Participants were recorded individually during a single session. A participant first completed a statement of informed consent and a questionnaire to gather biographical data and then was asked by the experimenter to mark his or her Indiana hometown on a state map. The experimenter and participant then moved to a small, sound-treated studio where the participant was seated in front of three navy blue baffles. A laptop computer was elevated to eye-level on a speaker stand and placed approximately 50-60 cm in front of the participant. Prompts were presented to the participant in a Microsoft PowerPoint presentation. The experimenter was seated directly next to the participant, but outside the camera angle, and advanced the PowerPoint slides at a comfortable pace. Participants were recorded with a Panasonic DVC-80 digital video camera to miniDV digital video cassette tapes. All participants wore a Sennheiser MKE-2060 directional/cardioid lapel microphone throughout the recordings. Each speaker produced a total of 94 segmented files which were converted from Final Cut Express to Quicktime (.mov) files and then saved in the appropriately marked folder. If a speaker mispronounced a sentence or word during the recording process, the mispronunciations were edited out of the segments to be archived. The remaining parts of the recording, including the correct repetition of each prompt, were then sequenced together to create a continuous and complete segment. The fourteen participants were between 19 and 61 years of age (with a mean age of 30 years) and native speakers of American English.

ESD is an Emotional Speech Database for voice conversion research. The ESD database consists of 350 parallel utterances spoken by 10 native English and 10 native Chinese speakers and covers 5 emotion categories (neutral, happy, angry, sad and surprise). More than 29 hours of speech data were recorded in a controlled acoustic environment. The database is suitable for multi-speaker and cross-lingual emotional voice conversion studies.

These carpete contains the datasets features used and described in the research paper entitled
García-Cuesta, E., Barba, A., Gachet, D. "EmoMatchSpanishDB: Study of Speech Emotion Recognition Machine Learning Models in a New Spanish Elicited Database" , Multimedia Tools and Applications, Ed. Springer, 2023

In this paper we address the task of real time emotion recognition for elicited emotions. For this purpose we have created a publicly accessible dataset composed by fifty subjects expressing the emotions of anger, disgust, fear, happiness, sadness, and surprise in Spanish language. In addition, a neutral tone of each subject has been added. This article describes how this database have been created including the recording and the performed crowdsourcing perception test in order to statistically validate the emotion of each sample and remove noisy data samples. Moreover we present a baseline comparative study between different machine learning techniques in terms of accuracy, specificity, precision, and recall. Prosodic and spectral features are extracted and used for this classification purpose. We expect that this database will be useful to get new insights within this area of study.

The first dataset is "EmoSpanishDB" that contains a set of 13 and 140 spectral and prosodic features for a total of 3550 audios of 50 individuals reproducing the 12 sentences for the six different emotions, ’anger, disgust, fear, happiness, sadness, surprise’ (Ekman’s basic emotions]) plus neutral.

The second dataset is "EmoMatchSpanishDB" and contains a set of 13 and 140 spectral and prosodic features for a total of 2050 audios of 50 individuals reproducing the 12 sentences for the six different emotions, ’anger, disgust, fear, happiness, sadness, surprise’ (Ekman’s basic emotions]) plus neutral. These 2050 audios' features are a subset of EmoSpanishDB resulting of the matched audios after application of a crowdsourcing process to validate that the elicited emotion corresponds with the expressed.

The third dataset is "EmoMatchSpanishDB-Compare-features.zip" that contains the COMPARE features for the experiments of dependent-speaker and LOSO. These datasets have been used in the paper "EmoMatchSpanishDB: Study of Machine Learning Models in a New Spanish Elicited Dataset" and their creation, its contents, and also a set of baseline machine learning experiments and results are fully described within it.

The features are available under MIT license and if you want to get access to the original raw audio files for creating your own features and research purposes you can get them under CC-BY-NC completing and signing the agreement file (EMOMATCHAgreement.docx) and sending it via email to esteban.garcia@upm.es

The STC Russian speech database was recorded in 1996-1998. The main purpose of the database is to investigate individual speaker variability and to validate speaker recognition algorithms. The database was recorded through a 16-bit Vibra-16 Creative Labs sound card with an 11,025 Hz sampling rate.The database contains Russian read speech of 89 different speakers (54 male, 35 female), including 70 speakers with 15 sessions or more, 10 speakers with 10 sessions or more and 9 speakers with less than 10 sessions. The speakers were recorded in Saint-Petersburg and are within the age of 18-62. All are native speakers. The corpus consists of 5 sentences. Each speaker reads carefully but fluently each sentence 15 times on different dates over the period of 1-3 months. The corpus contains a total of 6,889 utterances and of 2 volumes, total size 700 MB uncompressed data. The signal of each utterance is stored as a separate file (approx. 126 KB). Total size of data for one speaker approximates 9,500 KB. Average utterance duration is about 5 sec.A file gives information about the speakers (speaker?s age and gender). The orthography and phonetic transcription of the corpus is given in separate files which contain the prompted sentences and their transcription in IPA. The signal files are raw files without any header, 16 bit per sample, linear, 11,025 Hz sample frequency. The recording conditions were as follows:Microphone: dynamic omnidirectional high-quality microphone, distance to mouth 5-10 cmEnvironment: office roomSampling rate: 11,025 HzResolution: 16 BitSound board: Creative Labs Vibra-16Means of delivery: CD-ROM