Learning the Structure of Music

“Le StruM is a project that comes from three areas, Music Cognition, Representation, and Machine Learning. This project is aimed at the development of models and tools for the application of novel machine learning techniques to the analysis of music and brain imaging data on music cognition.”

This project is different than most projects in perception research, as that it does nothing to analyze anything visual, only audible. The project uses the techniques described above to be able to use a computer to be able to understand the “relationship between musical structure and musical performance” as well as the listening experience that we are so used to, but are not able to explain. Some papers being submitted to the website are also attempting to link the brain with the science of music perception, convinced that understanding the brain would unlock the doors to understanding the structure of music as we know it. Researchers from places like the Institute of Human Cognitive and Brain Sciences are using this idea to move forward on actually understanding the brain, through using music. In fact, a large swathe of the research being done in the past couple years have relied on using the history of music in our culture to attempt to understand how the brain functions, believing the way music accesses memory, visualization, and more hold the key to the brain.

Relevant Link: Le Strum

RoboCup

“By the year 2050, develop a team of fully autonomous humanoid robots that can win against the human world soccer champion team.”

This is the stated goal of the RoboCup project. This project is a loosely associated group of researchers from around the globe, attempting to build a robotic team of soccer players within the first half of the 21^st century. This project, since it involves robotics, will use every facet of current motion perception technology. At this point in robot development, creating a Roomba-like robot may not be too difficult to simply shove a soccer ball into the goalie’s net. However this competition is much more ambitious: to create humanoid robots that must adhere to standard soccer rules. Throwing aside the issue of what kind of soccer players would want to risk their legs against a robotic one in a slide tackle, it is an interesting project. Each year a few papers are published exclusively for the purpose of bettering the general knowledge in regards to the subject.

This year’s competition is being held in Atlanta, in early July of 2008. The registered teams for this year’s competition are diverse, including China, Brazil, Mexico, Iran, Japan, and more. In fact it appears that outside of the four-legged and middle-sized competition, the United States is not a major participant in the competitions. This could be due to a variety of factors, none of which I am qualified to speak on. Humorously, in the schedule section of the website for the current competitions, teams get a full “practice day” to help their robots get with the program. See the participation table for schools that are doing research.

Relevant Link: RoboCup

Interview with Professor David Jacobs

In an interview I performed with Professor David Jacobs of the University of Maryland, he said in the next five to ten years, major advances in the field will allow this field of computer science to break out into the mainstream consumer culture, and impact the lives of everyone. As a researcher of computer vision, he stated that computer vision is the study of trying to get information out of images, to understand the world through images in other words, and to understand motion. The techniques used in this field involve three-fold:

1. Analyze image using image processing.
2. Analyzing geometry and optimization thereof.
3. Machine learning using different types of training data.

At its face, these problems seem fairly abstract, but they are all an effort to overcome the challenges that still lie ahead in the field. Dr. Jacobs asserts that the largest obstacle to wide-spread use of this technology is that is it very difficult to get a relationship between two-dimensional and three-dimensional objects. The most interesting problem he is tackling right now is collaborating with the Smithsonian institute on tacking pictures of leaves and being able to identify species. This is done by attempting to analyze the difference in shapes between leaves, using silhouette processing. All techniques discussed above are being used to figure out an optimal solution.