Hungarian speech database for computer-using environment in offices

Klára Vicsi, András Kocsor, Csaba Teleki, László Tóth

Speech databases were recorded in different offices, laboratories, and homes. Recordings in all scenes were prepared by using two parallel synchronized recording systems. One of the recording systems is the so-called reference system, where a close talking microphone (Monacor EMC 100) and a good quality sound card (Hercules Muse Pocket USB 5.1) and a laptop (Gericom Webshox) were used. The second recording system was the so-called varied system, where different microphones, sound cards and PCs were applied.

Description of the database:

The recording form was 16 bits and 16 kHz

ˇ        332 speakers;

ˇ        Twelve sentences and twelve words per speaker from a phonetically balanced text, composed in accordance with special Hungarian phonetic expectances;

ˇ        Large variations of different microphones, sound cards and PCs were used;

ˇ        Computer-using environment in offices, homes and laboratories;

ˇ        The whole material of database is annotated and one third (100 speakers) is segmented and labeled by hand.