Rapid porting of ASR-systems to mobile devices

Abstract

Portable devices for the consumer market are becoming available in large quantities. Because of their design and use, human speech often is the input modality of choice, for example for car navigation systems or portable speech-to-speech translation devices. In this paper we describe our work in porting our existing desktop PC based speech recognition system to an off-the-shelf PDA running WindowsCE3.0. We do this in a way that our already well performing language and acoustic models can be taken over without the need of retraining them for the PDA. In order to achieve an acceptable run-time behavior we apply several optimization techniques to the preprocessing and decoding process. Among other things we introduce the newly developed early feature vector reduction. In that way the execution time of our recognition system can be reduced from initially 28x realtime to 2.6x real-time with a tolerable increase in word error rate. The size of the acoustic models is reduced to 25% of its original size.

Optimizing components for handheld two-way speech translation for an English-iraqi Arabic system

“…As the main input modality, we implemented speaker independent speech recognition using the integrated microphone of the mobile device. The ASR system uses the Janus Recognition Toolkit (JRTk) featuring the IBIS decoder [3] for the laptop version, and a PDA-optimized port [4] for the PDA version, that runs around 2-5x real-time on the 624MHz Intel XScale PXA270 processor. This is the standard processor found in most commercial PDAs.…”

“…The Early Feature Vector Reduction [4] (EFVR) is used to remove redundant consecutive feature vectors, as found in silence and static vowels or noises, which results in a reduction of 25 to 50% of the feature vectors before they are fed into the decoder.…”