Distributed speech recognition using phoneme lattices
Hagai Aronowitz
In: ICASSP 2007, 15-20 April 2007, Hawai, USA.

Abstract

In this paper we suggest using phoneme lattices as an interface between a client and a server in a distributed speech recognition framework. According to this approach, the client should perform all the computation that is tightly coupled to the user's acoustic characteristics, and the server should perform all the computation that involves lexical and grammatical knowledge. We present an algorithm that aims to both improve accuracy of such a system and minimize communication complexity. Our algorithm can also be used for audio indexing for keyword retrieval.

PDF - Archive staff only - Requires Adobe Acrobat Reader or other PDF viewer.

[Edit]