MusicRainbow Demo

I finally finished the video for the MusicRainbow demo. It's only 3'20'' long but it took almost 5 years to make it (including at least 7 years waiting for the program to encode the different video formats). At least I learned a lot about codecs and which ones are not suitable for this kind of video. I'll post it to Information Aesthetics as suggested by Paul (thanks for the suggestion!) :-)

I built the interface with Processing. It's a wonderful tool to quickly develop ideas and one of the many great suggestions Masataka has given me in the last 7 months. Currently I'm working on another user interface using Processing... which I think will be much better, because it takes user preferences into account such as "I like this artist" and allows exploration within different dimensions... and first results seem great too :-)

BTW, Masataka is a master in creating demonstration videos. You might be interested in some of them. For example, in one of his demonstration videos he shows how we will scan through songs in the future (SmartMusicKIOSK). Another futuristic tool with nice videos is Musicream which can be used to generate playlists and explore music.

CrestMuse & SIGMUS

On Thursday we had a CrestMuse project presentation in Osaka. (That’s the project I’m working on.) Several partners are involved who work on a wide range of very interesting projects related to music. Somehow they managed to get over 100 people to attend (and I was told that they didn’t do much advertisement). They even had to reject a large number of students because they ran out of seats in the lecture room.

The meeting took place on the 14th floor of a stylish skyscraper. I would have never guessed that we were actually in university offices. The location was great (right in the city center). The view of the city was great too. However, the strangest thing was that the university was sharing a building with a hotel (which looked really nice), and a fancy restaurant. (And probably a number of companies.) Masataka told me that he thinks “normal” students don’t study there, but he wasn’t sure what kind of students actually do.

View of Osaka from one of the rooms where the demonstrations were given.



The project presentation consisted of lots of demonstrations (including stuff I’ve been working on), some overview talks, and a panel discussion. I couldn’t understand the Japanese presentations, but the demonstrations were great, and the best part was to meet many colleagues I had already met in summer. I was also told that the main purpose of such meetings was to get drunk :-)

After going out for some drinks, I spent most of the remaining night writing slides and restructuring them for my invited talk for the next morning… The next day a SIGMUS meeting took place in the same rooms. SIGMUS is the Japanese special interest group for music. They have several meetings a year in different locations in Japan, the next one (which will be even bigger) is in December in Tokyo. I was impressed how large and how well organized the music research community is in Japan. It seems like a parallel world to the “international community”. Unfortunately many papers never get translated into English. The language barrier is definitely a problem… but the situation seems to be improving. Anyway, the demonstrations were great. Unfortunately, the schedule was very tight and there were lots of people… so I didn’t get to play with the demonstrations as much as I wanted to. I can’t wait until such music interfaces will be available to everyone :-)

This pictures was taken in one of the rooms were the demonstrations were given. There was music from all directions :-)



In the front of the next picture you can see Yoshii's Drumix demonstration. I really enjoyed playing with it! You can put any audio CD into the system, it automatically analyzes the bass and snare drum parts, lets you replace them with other samples, lets you change the rhythm patterns, ...
In the background you can see Hamanaka who also demonstrated his work at ISMIR. Unfortunately the speakers he used at ISMIR didn't make it possible to really understand what his demonstration was about. Anyway, I got another chance to try it out (with headphones) and it was really fun... it takes a bit to get used to how to control the system, but it's fun to play with. Standing in the middle of the demonstration room I could hear different music from all directions. Moving my hand towards my ear I could focus on the music coming from in front of me.



There were lots of other cool demonstrations which I don't have pictures of. I especially liked Ito's work on remixing a piece of music. It was fun to control the loudness of each individual instrument. I think it's a great way for untrained listeners to understand how each instrument contributes to the whole. Ito explaining his system:



I also liked the demonstration by Kitahara who found a clever way to compute similarity on the instrumentation level. I think his approach could help overcome limitations the techniques have which I'm currently working with.

Tommie's demonstration on the evaluation of singing skills was lots of fun, too. Unfortunately his system only gave me 30 out of 100 points... it thought I didn't get my pitch intervals right. I think he still needs to improve it.

There were many more interesting demonstrations. For example, Ohishi demonstrated a nice query-by-humming and speech interface, and Prof. Katayose's team also had a number of very interesting demonstrations which were fun to try out. I also had the opportunity to have interesting conversations with many people, including Keiji Hirata.

Goto's Lab

Masataka Goto is supervising my work at AIST. At the same time he is supervising at least 7 other students at the lab. Most of them are spending one or two months of their summer vacations at AIST (like an internship).

I've seen various styles of supervision of students, but I've never seen anyone so intensively supervising so many students at the same time. Sometimes it seems that he has meetings with students all day... one day after the other. All students have very different topics. And all of these topics are really interesting. None of the students needs to worry that another student could "steal" an idea from them, or come up with a solution to the problem they are working on before they do. Instead they identify areas where results from one project could help another one, and work together. Every student also seems to have a very clear picture of what they are working on, what the next steps could be, and how they want to evaluate progress. It's amazing!

I used to think that the best way to work together in research is to be working towards the same goal. I guess that can work really well in special cases, but it is somehow a problem when you need to write two different doctoral dissertations about the same topic, or when it comes to the order of authors on a publication. However, given this experience in Goto's lab I think the best way (in computer science) is to work on different but somehow related topics and just sit next to each other. (Most of the students' topics deal with music analysis in some way.) However, this also means that one needs to come up with many interesting topics to work on...

BTW, today I learned that in Japan the labs at universities are officially named after the professors. Instead of names such as "Intelligent Music Processing Group", "Music Technology Group", or "Centre for Digital Music", the last name of the professor is used. In general it seems that professors are very highly respected. However, it seems a bit strange to me that the name of the professor is more important than the research field the group is actually working on.

Here is a picture of Masataka at his desk. He actually did some cleaning before I was allowed to take a picture... but I didn't notice any difference :-)



These are the students Masataka supervises (including Masataka, Doug, and me).



This picture was taken a few hours later at the "all-you-can-drink" place. (Yes, alcohol!)



And this is my new business card in Japanese! :-)



Business cards are very important in the Japanese communication culture. If I would give you one of my cards, you should receive it with both hands. Look at it, marvel at its beauty (or talk about how impressive the font is) and then after 10 minutes or so of admiring it, carefully put it into the special shiny metal case in which you store all the important business cards you receive. Do not take notes on the card! And do not stick the card in your back pocket! The way you treat the card is an indication of what you think about me :-)

Here are two updates to my last posts. First, here you can see what "PreKla" looks like when experts do it. And here (2.45MB) you can find a great video of the HRP2 robot (which is partly developed by some of my colleagues at AIST) dancing the dance which I saw the old ladies dance last Sunday.



Find out more about this dancing robot here (in Japanese).