Design of a voice-enabled interface for real-time access to stock exchange from a PDA through GPRS

In this paper we describe the design of a voice-enabled interface embedded in a Personal Digital Assistant (PDA). It is a component of a more complex system which allows the access to real-time information (for example, prices on stock exchange) via the General Packet Radio Service (GPRS). Specifically, we have focused on the Automatic Speech Recognition (ASR) module: we have implemented a fixedpoint version of the whole ASR system and recognition results showed that its performance is very similar to that achieved with its floating-point implementation. In addition, with the purpose of facilitate as much as possible the portability of the ASR system to other hand-held devices, we have dealt with the problem of adapting a set of general acoustic models to match a specific acoustic environment when few adaptation data is available.