By Brookes S.D.
Read Online or Download A Model for Communicating Sequential Processes [PhD Thesis] PDF
Best nonfiction_6 books
- Perfect, amicable, and sociable numbers : a computational approach
- Crystallization and Mtls. Sci. of Mod. Artificial, Natural Crystals
- Modesty Blaise 09 - The Head Girls
- Dark Matter Search [lg article]
- Central Electronics 200V
- Australia - Western Australia - The Big Picture
Additional info for A Model for Communicating Sequential Processes [PhD Thesis]
Promising work has been presented to take advantage of the correlation between more holistic body motion and speech [31, 32, 59, 60]. Such methods have shown a relationship between global body motion and speech over longer term sequences. The experiments presented in this chapter, continues in this direction, exploring the extent to which we can use findings in the psychology literature to address the audio-visual clustering problem in meetings more directly for constructing a plausible practical approach to the problem of speaker localization.
7, we describe a method of associating audio-visual data and present bench-marking results. We conclude in Sect. 9 and discuss the future challenges. 2 Background Clustering audio-visual meeting data can involve the grouping of events on different levels. From the coarsest level, we may want to group them based on date, location, or which work-group participated. If we increase the granularity, we observe events within a single meeting such as the types of activities that took place. Increasing the granularity further, each activity consists of a conversation type (ranging from monologue to discussion) where speech turn-taking events occurs.
In more natural meeting scenarios, people do not cut from doing a presentation to a discussion or a monologue necessarily so annotating these meetings in terms of meeting actions is not practical. With this in mind, it is probably easier to extract semantically meaningful features which are easier to evaluate. The problem with analyzing meeting actions is that labeling is strongly dependent on the temporal context. Rather than examining temporal intervals of time, we can also segment based on events such as a change of speaker or when someone starts or stops speaking.
A Model for Communicating Sequential Processes [PhD Thesis] by Brookes S.D.