@misc {pub2516,
	title = {Steps towards a more natural and intuitive communication between humans and machines
},
	author = {Martin Ernst Heckmann},
	year = {2013},
	month = {November},
	abstract = {Current speech interfaces are perceived as little intuitive and error prone. In this talk I will present different strategies to make the human-machine interface more intuitive and less fragile by enriching the range of information available to the machine and tolerating changes in the acoustic environment. This includes features more robust against noise and approaches to capture prosodic variations.

I will start with the hierarchical spectro-temporal features we developed. Inspired by findings on neuron responses in the mammalian cortex these features jointly extract spectro-temporal information. This is particularly beneficial in noisy conditions when combined with conventional spectral features.

Next, I will briefly present algorithms we developed for the extraction of the formants and the fundamental frequency.

Finally, I will talk about my recent experiments on the audio-visual extraction of prosodic cues. Such prosodic cues are known to play a very important role for human-human communication, e.g. they differentiate between questions or statements or emphasize certain words (e.g. novel information) in an utterance. Yet they are rarely used in the man-machine interface. The target scenario is the detection of a spoken correction of a user. In this case the corrected word is expected to be highly prominent. I will on one hand demonstrate how prominent words can be discriminated from non-prominent ones. On the other hand I will show that the information on the mouth shape is a very strong cue for many speakers, yielding similar or better results than e.g. fundamental frequency.},
	publisher = {Leibniz Zentrum f{\"u}r Informatik},
	booktitle = {Dagstuhl Seminar: Computational Audio Analysis, Schloss Dagstuhl},
	city = {Dagstuhl},
	address = {Dagstuhl},
	institution = {Leibniz Zentrum f{\"u}r Informatik}
}