Duration: 1:25:38

111 views

1

As the prevalence of SMS/IM increases on cell phones and other mobile devices, the need for efficient and robust text entry interfaces increases. By combining speech and keypad inputs, the SpeechPad prototype application demonstrates the design of a multimodal interface that improves text entry rate on keyboard-less mobile devices. The use of the familiar per word entry interface increases the perceived accuracy and hides recognition latency. Furthermore, the use of an efficient T9-based filtering of recognition lattice and n-gram results achieves graceful degradation as recognition accuracy decreases. In this talk, we will examine the designs, algorithms, and implementation considerations of this multimodal text entry interface. We will also show the Pocket PC SpeechPad prototype that some have labeled as the next killer app for speech recognition.