Hello and welcome to this site, which is dedicated to the author's interests in Audio Visual Speech Recognition (AVSR) and, more generally, in Automatic Speech Recognition (ASR).

Information and resources that were produced developping some recognition prototypes can be found in Projects.

Some didactic material is available in Education.

Basic information, including contacts, concerning the author are available in About the author.




      The ABCP-1 system is a speaker dependent, large vocabulary and continuous speech recognizer, that is based on acoustic and visual speech features.

      Follows some of the available technical reports (and respective slides):

  • ABCP1-TR-00 (slides-00) - a general overview of the recognizer;
  • ABCP1-TR-01 - the ABCP-db1 application specific corpus;
  • ABCP1-TR-02 (slides-02) - the system's language model
  • ABCP1-TR-03 - the Acoustic Model;
  • ABCP1-TR-04 - the Visual Model;
  • ABCP1-TR-05 - the Decoder;
  • ABCP1-TR-06 - the Acoustic Acquisition and Analysis Module;
  • ABCP1-TR-07 - the Visual Acquisition and Analysis Module. (Date modified: 02/07/2014

      A few notes on building a DBN-based ABCP-1 version (WORK IN PROGRESS). 

      Other technical reports: ABCP1-TR-14-01 - "A New Method for Language Modeling". (WORK IN PROGRESS - Date modified: 18/09/2014)

      From the Language Model of the ABCP-1 recognizer, some Lexicons, nGrams and Perl scripts are available in this directory

      In relation to the the initial version of theVisual Acquisition and Analysis Module, are also available the C++ code and a sample video (23MB AVI or 500kB FLV).



      This project has two main goals: 1) to gain some insight on (Human) Audio Visual Speech Perception (AVSP); 2) and to produce a brief survey on the AVSP state of the art from an AVSR perspective (WORK JUST INITIATED).



A primer on Automatic Speech Recognition (portuguese language only) is available.


About the Author


My name is Vitor Pera, I'm Portuguese (born in Angola, 1963).

I'm Engineer, I've specialized in Automatic Speech Recognition (PhD on "ASR multistreaming", FEUP 2001).