Music Collage (music information retrieval, human computer interaction)
| Omar Al-Hasani (PM) | ||
| Boris Pfahringer | ||
| Jeremy Roundill | ||
| Brian Compfers |
Collaging is a technique developed by Andruid Kerne where images and fragments of text are displayed over time at 'random' positions on a canvas. Sources for these images and text are typically from external web sites -- for instance you might start a collage from the home page of the BBC, CNN, and Fox News. The program them starts to crawl these sites, displaying images from these sites and excerpts of text. When a user sees an item of interest they click and are taken to the page that item is from.
The aim of this project is to apply this idea to audio -- the project will require a technique for associating images with audio (such as audio fingerprinting) -- enriched with the idea of allowing user interaction to indicate which items a user likes (i.e. show me more), and which ones they do not like -- the project will also need a way to automatically genre/mode classification.
Such audio analysis techniques have been researched for over a decade through the topic of music information retrieval, and there is a healthy supply of open source software implementing a variety of techniques.
Play me Music! (mobile devices, audio processing)
The aim is simple: when travelling by car, develop a plug-n-play device that will always let you listen to music on the radio
Suggested approach: programme a mobile device (e.g. an iPod touch) to monitor two radio stations (stereo split mic in), play one station and buffer the other. On detection of speaking (i.e. not music) switch to other buffer. Repeat process. Output piple to mini radio transmitter that is then picked up by car radio
Party Pepper/Cocktail Smarty (digial signal processing, natural language processing, alerting system)
| Al Al Bushi | ||
| Mark Will | ||
| Matt Hunter | ||
| Michael Fowke | ||
| Brett Saddleton |
Bored at a party? Embarassed by lull in the conversation? Then the answer to your problems is here with Party Pepper. The aim of this project is to develop a software program that monitors what people are saying/talking about, and retrieve relevant information from the web and display it on a (large-ish) screen.
Keyword spotting through speech recognition, linked to media display/playback (visual: photos, video?)
Cocktail Smarty would be a variation on this, targetted at more organized/form occassions (such as a conference dinner), perhaps primed to act as a live who's who's (distributed, microphones at each table, connect topics?)
Digital Library back-chat -- Co-Lib (digital libraries, network protocols)
| Craig Robb (PM) | ||
| Eru Penkman | ||
| Steve McTainsh | ||
| Nathan Holland |
Digital libraries that "talk" to one another to exchange information
Surprisingly, this isn't a capability that many (any!) existing digital library systems are capabable of, despite over a decade of research in the field. This lack of this ability is all the more bizare when one considers that the physical equivalent has been possible for centuries. When walking into a bricks-and-mortar library, if they don't have the item you are looking for they will be able to locate and then borrow/copy the item from another library.
Well, that's all about to change with this project!
The origins for this project came from a real world projet where they are looking to use video Nano iPods quite literally "in the field" to capture videos, managed as a digital library. When docked with a networked PC, documents are synchronized with central DL server. Developing a mechanism for the two digital libraries to 'chat' to one another would provide a framework in which to implement this -- and more importantly, provide a more general mechanism that could be applied to a much wider variety of problems.
See Greenstone for more details about the DL software developed at Waikato and used by the United Nations, BBC, National Library of NZ, ...
Symphony (visualization, video analysis, cluster computing)
| Han Byeol Son (PM) | ||
| Reuben Bell | ||
| Anthony Coddington | ||
| Ethan Duff |
The challenge: to develop a compelling visualization for a large scale display (v. high resolution) is S-block
User interface: Hand gestures captured by cameras.
One possible idea based around pannable, zoomable world map, showing:
geo-located twitters in real-time
visualizations for world census data or other globally collaged statistics (e.g. geo.worldbank.org).
....
Other themes includes a blended montage/collage from homogeneous (but disparate) sources around the world, or a mash-up of heterogeneous sources. An example of the former would be blended CCTV from around the world; an example of the latter would be to take keywords from a variety of news stories and use then to locate images on a photo site such as Flickr. In both cases, again there would be the opportunity for user to interact with the visualization, and direct where and how the feeds alter over time.
ManyEyes gives an excellent gestalt of suitable visualization techniques.
Top Gear Electrified (3D Graphics, Physics Enginges)
| Krishneel Raj | ||
| Siva Manoharan | ||
| Adrian Shaw | ||
| Siddharth Arora | ||
| Martin Poot |
Countless kids have had hours of fun playing with Scalextrics or similar car racing games. This project is about building a digital version of this: building the track; having a range of cars (truck?) etc. to choose from; and running them on the track. Not only can the basic physical modeling of the environment be done, but being in the digital domain extra magic is possible. There can be limitless pieces if you want, and special operations such as the "automatically join these two pieces." And that's just the beginning ...
Things to think about:
Oh yes, and it means grown-ups can play at toy cars without having to make a public spectacle of themselves.
Expediteee: Parallel Programers/Collaborative Programming with Spatial Hypermedia (expeditee, networks, protocols, versioning control)
| Benson Chang | ||
| Brad Cowie | ||
| Jared Lindsey | ||
| Luke Barnett | ||
| Tataihono Nikora |
Skypeless (networking, dsp)
Seed of idea (suggestion from class) started as Skype without Skype. Voice over IP without needing to install an application. Broader idea to fit it into Expeditee (new generation of Web browser, demo'd by Rob) where people collaborate (visual and audio) when viewing the same frames in Expeditee.
Apply operation transforms (see Google Wave) to the underlying frame file format of Expeditee???
Distributed Expeditee (expeditee,distributed software)
Data transfer based around UDP rather than TCP