SHACHI - Language Resource Metadata Database

Desktop/Microphone
8 hours of speech as spoken by 2 female speakers and 2 male speakers for each language (English and Spanish).

Telephone
The LILA Marathi database collected in India was recorded within the scope of the LILA II project. It contains the recordings of 2,002 Marathi speakers (992 males and 1010 females) recorded over the Indian mobile telephone network.

The following acoustic conditions were selected as representative of a mobile user's environment (some speakers were recorded in several environments):
- Passenger in moving car, railway, bus, etc. (153 speakers)
- Public place (285 speakers)
- Stationary pedestrian by road side (237 speakers)
- Home/office environment (945 speakers)
- Passenger in moving car using a hands-free kit (382 speakers)

This database is distributed as 2 DVD-ROMs. The speech files are stored as sequences of 8-bit, 8kHz A-law speech files and are not compressed, according to the specifications of LILA. Each prompt utterance is stored within a separate file and has an accompanying ASCII SAM label file.

This speech database was validated by Appen (Australia) to assess its compliance with the LILA format and content specifications.

Each speaker uttered the following items:
- 2 isolated digits
- 6 connected digits (1 sheet number -5+ digits, 2 read telephone numbers –9/11 digits, 1 credit card number –14/16 digits, 1 PIN code -6 digits, 1 spontaneous telephone number)
- 1 natural number
- 1 currency money amount
- 3 dates (1 spontaneous date e.g. birthday, 1 word style prompted date, 1 relative and general date expression)
- 2 time phrases (1 spontaneous time of day, 1 word style time phrase)
- 3 application words
- 1 spotting phrase using an embedded application word
- 5 directory assistance names (1 spontaneous, e.g. own surname, 1 city of birth/growing up, 1 most frequent city out of a set of 500, 1 most frequent company/agency out of a set of 500, 1 “forename surname” out of a set of 150 )
- 3 spelled words (1 surname, 1 directory assistance city name, 1 real/artificial name for coverage)
- 1 “silence word”
- 13 phonetically rich sentences
- Min. 6 spontaneous tasks of 10 to 15 seconds of speech

The following age distribution has been obtained: 878 speakers are between 16 and 30, 815 speakers are between 31 and 45, and 309 speakers are between 46 and 60.

A pronunciation lexicon with a phonemic transcription in SAMPA is also included.

Desktop/Microphone
A-SpeechDB© is an Arabic speech database suited for training acoustic models for Arabic phoneme-based speaker-independent automatic speech recognition systems. The database contains about 20 hours of continuous speech recorded through one desktop omni microphone by 205 native speakers from Egypt (about 30% of females and 70% of males), aged between 20 and 45.

Automatically generated transcriptions are provided with a manually revised version for each sentence.

Accuracy:
• Detailed speaker information: Age, Accent, place of stay, gender
• Recording in office environment
• Sentence labeled.
• Continuous Speech
• Automatic first pass transcription
• Manual second pass labeling
• Each text prompt is unique, no repeated sentences
• Sentences chosen to cover all Arabic phonetics several times

Technology:
• Automatic transcription using TransArab©
• Recording using DBRec© or Validator©
• Validation using Validator©

Audio:
• Sample Rate : 16 KHz
• Resolution: 16 bit PCM
• Format: MAF (A tool is included to convert the database to WAV format)

Text:
• Labeled data format: HTK lab format (100 nano-seconds)

Telephone
This speech database contains the recordings of 2000 Catalan speakers who called from Fixed telephones and who are recorded over the fixed PSTN using and ISDN-BRI interface. Each speaker uttered around 50 read and spontaneous items. The speech database follows the specifications made within the SpeechDat (II) project. The database was validated by UVIGO. The Catalan-SpeechDat for the Fixed Telephone Network Database was funded by the Catalan Government.

Telephone
This speech database contains the recordings of 2000 Catalan speakers who called from GSM telephones and who are recorded over the fixed PSTN using and ISDN-BRI interface. Each speaker uttered around 50 read and spontaneous items. The speech database follows the specifications made within the SpeechDat (II) project. The database was validated by UVIGO. The Catalan-SpeechDat for the Mobile Telephone Network Database was funded by the Catalan Government.

Desktop/Microphone
This database contains parliamentary statements and newspaper reviews read by a semi-professional male speaker. It consists of a selection of 2150 sentences annotated and manually verified, including 100 rare phonemes in words. Prompts vary in length from 2.3 to 13.4 seconds, with an average length of 6.3 seconds.

The recordings took place in an anechoic chamber using one table stand dynamic microphone (Rode NT1000). A 48 kHz sampling frequency and 16 bit resolution was used. The total duration of the recordings is 3.45 hours.

The signal was automatically aligned with the transcription, and manually corrected using Praat speech analysis program. The database is phonetically annotated and manually corrected, which represents a lexicon of 11761 words with phonetic transcription.

The package also includes a version of the speech database re-sampled at 16 kHz and edited. In all these files DC offset and the identified distortions which could affect the quality of speech synthesis were removed using High-pass filter.

For a more detailed description, see “Oliver D. Szklanny K. Creation and analysis of a Polish speech database for use in unit selection synthesis, LREC Genoa, Italy 2006”: http://www.lrec-conf.org/proceedings/lrec2006/pdf/688_pdf.pdf

Desktop/Microphone
The Catalan SpeechDat-Car database contains the in-car recordings of 300 speakers who uttered from around 120 read and spontaneous items. Each speaker recorded two sessions. Recordings have been made through 4 different channels, via in-car microphones (1 close-talk microphone, 3 far-talk microphones). The 300 Catalan speakers were selected from 5 different dialectal regions and are balanced in gender and age groups. The database was validated by UVIGO. The Catalan-SpeechDat-Car Database was funded by the Catalan Government.

Desktop/Microphone
The Catalan Speecon database comprises the recordings of 550 adult Catalan speakers who uttered over 290 items (read and spontaneous). The data were recorded over 4 microphone channels in 4 recording environments (office, entertainment, car, public place). The speech database follows the specifications made within the UE funded Speecon project. The database was validated by UVIGO. The Catalan-Speecon Database was funded by the Catalan Government.

Desktop/Microphone
EUROM1 is a multilingual European speech database. It contains over 60 speakers per language who pronounced numbers, sentences, isolated words ... using close talking microphone in an anecoic room. Equivalent corpora for each of the European languages exist already, with the same number of speakers selected in the same way, and recorded in the same conditions with common file formats.

Desktop/Microphone
This database contains the recordings of one male and one female Spanish professional speakers recorded in a noise-reduced room. It consists in recordings and annotations of read text material in neutral style plus six MPEG expressions, all in fast, slow, soft and loud speech styles. The text material is composed of 184 items including phonetically balanced sentences, digits and isolated words. The text material was the same for all the modes and styles, giving a total of 3h 59min recorded speech for the male speaker and 3h 53min for the female speaker. The Emotional speech synthesis database was created within the scope of the Interface EU funded project.