Language model combination and adaptation using weighted finite state transducers

In speech recognition systems language model (LMs) are often constructed by training and combining multiple n-gram models. They can be either used to represent different genres or tasks found in diverse text sources, or capture stochastic properties of different linguistic symbol sequences, for example, syllables and words. Unsupervised LM adaptation may also be used to further improve robustness to varying styles or tasks. When using these techniques, extensive software changes are often required. In this paper an alternative and more general approach based on weighted finite state transducers (WFSTs) is investigated for LM combination and adaptation. As it is entirely based on well-defined WFST operations, minimum change to decoding tools is needed. A wide range of LM combination configurations can be flexibly supported. An efficient on-the-fly WFST decoding algorithm is also proposed. Significant error rate gains of 7.3% relative were obtained on a state-of-the-art broadcast audio recognition task using a history dependently adapted multi-level LM modelling both syllable and word sequences. ©2010 IEEE.

Enhanced efficiency of near-UV emitting LEDs for solid-state lighting applications

The performance of a series of near-UV (∼385 nm) emitting LEDs, consisting of high efficiency InGaN/AlInGaN QWs in the active region, was investigated. Significantly reduced roll-over of efficiency at high current density was found compared to InGaN/GaN LEDs emitting at a similar wavelength. The importance of optical cavity effects in flip-chip geometry devices has also been investigated. The light output was enhanced by more than a factor of 2 when the lightemitting region was located at an anti-node position with respect to a high reflectivity current injection mirror. A power of 0.49 mW into a numerical aperture of 0.5 was obtained for a junction area of 50μm in diameter and a current of 30 mA, corresponding to a radiance of 30 W/cm2/str.