Aholab Signal Processing Laboratory

Welcome to Aholab

Details

Written by Inma Hernaez Rioja

Aholab is the short name of the Signal Processing Laboratory of the University of the Basque Country (UPV/EHU). The laboratory is located in Bilbao, We are a university research team and  focus our research in the areas of Text to Speech Conversion, Speech and Speaker Recognition, and Speech Processing in general. Since 2005 we are a recognized research group of the Basque Research Network. The laboratory belongs to the Department of Electronics and Telecommunications (DET) in the Faculty of Engineering of Bilbao (ETSI). The group is dedicated to the speech processing technologies and its applications, with focus in the following specific areas:

• Text to Speech Conversion: The group has its own multilingual text-to-speech conversion system, working on English, Spanish and Basque. Our AhoTTS system for Basque (aholab.ehu.es/TTS) is the only one fully developed at the Basque Country and it is freely available. It is also able to generate emotional synthetic speech.
• Speech Synthesis: A big research effort is dedicated to the synthetic speech generation algorithms and technologies, to be incorporated into the AhoTTS for its evaluation. Virtually all state of the art speech generation technologies have been developed and evaluated through the last 15 years. At present we maintain a collaboration with the CSTR (Edinburgh) in order to include our last synthesis engine (AHOCoder) into their FESTIVAL speech synthesis tool.
• Prosody modelling: Prosody models and prosody generation techniques have been developed specifically for the Basque language. Special focus was on prosody analysis and conversion techniques applied to the generation of emotional speech.
• Speech recognition: The group has developed several public speech databases for the development and test of speech recognition systems for Basque, all of them available through ELRA. A reduced vocabulary isolated and connected word recognition system for Basque has also been developed, and a continuous speech recognizer is now being developed.
• Speaker recognition and speaker diarisation: We have participated in the past in several national projects involving speaker recognition. Presently this is a very active research area, with 2 live national projects and several international collaborations (see publications 2010). The recognition of the speaker emotions has also been a very active and productive research field during the last 4 years.

AHOcoder

Details

Written by Inma Hernaez Rioja

Ahocoder is a new speech coder very useful for statistical speech processing tasks. You will find useful information about it in AHOcoder download webpage