Monthly Music Hackathon rides again on August 5 and 6. The theme this month is Science of Music. The event takes place at Spotify’s NYC offices.
Some of the speakers took the time to chat about their work, about science, about music, and about all the interactions between the two. Finn Upham will talk about her work with music and breath. Maria Panteli will discuss using science to analyze music from around the world. Kaustuv Kanti Ganguli and Sankalp Gulati will speak about raga recognition.
What is Music Science? Why would you want to do science to music?
Kaustuv Kanti Ganguli: Bruno Mars has rightly said, “Music is not math. It’s science. You keep mixing the stuff up until it blows up on you, or it becomes this incredible potion.” The way I understand it is that music creation has a strong underlying mathematical model, but ultimately is an experimental paradigm. As long we discuss about music creation, there arises multidisciplinary perspectives such as musicality, artistry, sense of proportion, thoughtful use of ornamentation and many more. But underneath all of these, lies a concrete structure of the musical scale which is the backbone of the musical piece. I’d rather focus on this acoustics aspect of the musical scale (and instruments) as Music Science. This conforms with the necessary properties of science such as being measurable, quantifiable, repeatable and so on. Another realm of Music Science which is of prime importance is the perception and cognition of musical sounds. This viewpoint addresses very relevant research questions like music preference, listening, and pedagogy.
Music Science acts as a basis for any mathematical analyses on the music per se, may it be melodic or rhythmic analysis, study on timbral texture, structural segmentation of a music piece etc. Music Science is also an inevitable element in the study of systematic musicology where we aim to explore an objective definition of the music concepts and in the study of music psychology the goal is to find a musical memory representation or to discover the perceptual space of musical variabilities. Thus in my opinion, doing science to music should not be a choice but a requisite for any researcher in the field of engineering approaches to music applications. In any analysis-by-synthesis model of music exploration, Music Science is explains the rationale behind the concept and hence is a must do.
Kaustav Kanti Ganguli – Photo by Vikram
What are some basic things that music science is capable of that most people might not know about?
Maria Panteli: Algorithms can recommend songs you like but never listened to before or identify this melody that is stuck in your head or this sample that you hear looping. But music science is not just that. If you are a musician, it can help you explore unlimited sound effects or even act as your fellow musician, accompanying you and improvising along your lines. It can help in analysing a music piece by transcribing its chords, notes, rhythms and instruments or zoom in details of ornamentation and expressivity. But what I find most interesting about music science is that it can be applied to uncover trends over the years or capture evidence of cultural exchange. Because music, just like any other social interaction, evolves and it is exciting to be able to trace its steps as it goes around in the world.
Figure 2 From Maria Panteli’s paper: “Modeling Rhythm Similarity For Electronic Dance Music – Ismir 2014.”
Do types of companies other than music streaming companies use music science? How? Why?
Sankalp Gulati: The digital music consumption market has definitely influenced the research and technological advancements in the field music information research. However, music technology goes much beyond being utilized by the digital music distribution channels such as online music streaming. Music technology has been used in digital audio workstations for music production since a long time. Products such as Ableton Live, Logic Pro, Adobe Audition etc have been using music technology for manipulation and synthesis of sounds and music. It is now increasingly being used in several entertainment centric products such as karaoke and music instrument apps, for example, Sing Karaoke and Magic Piano (Smule). With an increasing infusion of machine learning in music technology, it has also been put to use for facilitating music education. For example, the Yousician app for learning instruments such as guitar, piano etc. and Riyaz (MusicMuni Labs) for learning vocals in Indian art music.
Sankalp Gulati – Photo by Shefali Bajpai
What is the biggest difference between what music science tells us music is, and what traditional knowledge tells us music is?
Finn Upham: The study of music cognition, an important part of music science, regularly challenges what we come to think of as music in our own day to day lives. From specific results of what the usual music listener can remember from the first minute of a piece, to clues of extra activity in the amygdala for one song vs. another, it’s our business to interrogate folk psychology. These days, my own definition of music is “a broadcast signal enabling sustained concurrent action” as this seems to be the criteria for engaging our musical perceptual processes, even when the signal can’t be heard. Our brains automatically use perception-action networks to follow and predict how such a signal should continue, with structural expectations informing our judgements of how we and others act along with it. These days, a lot of great research in music cognition focuses on to how we make and relate to music as groups of people, as children in families, putting it back into the context of a cultural phenomenon supported by our individual faculties, preferences, and goals. Many musical traditions have always treated music as a social activity, but some seem to have lost this perspective along the way to giving single consumers relatively convenient experiences.
Finn Upham – Photo by Elizabeth S.C. Wu
Science of Music Hackathon
On Friday, August 5, talks from DSP (Digital Signal Processing) wizards, cognitive scientists, and more start at 7:00 pm, followed by brainstorming for a full day of hacking on Saturday.
Saturday starts at 10:00 am, (also at the Spotify NYC offices) and will feature tutorials about libRosa, TensorFlow, the Amazon Echo, and ACRCloud, in parallel with the hacking. You can build anything you want related to music, science, and more. Original art, research, apps, and any other type of project are all fair game. Programming and technical skills are not required. Lunch will, of course, be served at noon. Demos of all the day’s hacks will start at 8:00 pm.
Spotify NYC
45 W 18th St, 7th Floor
New York, NY 10011
Free. All are welcome: RSVP
About Monthly Music Hackathon
Monthly Music Hackathon is a unique event series that brings together diverse NYC music communities to explore music from every angle. Once a month, an entire day is dedicated to a topic related to music, such as algorithmic composition, lyrics, hip hop, music games, and more. It’s an opportunity for you to learn about a new subject from experts, participate in hands-on workshops, work on your own project for a whole day, practice the full lifecycle of a creative project, and get feedback on your work from people with diverse viewpoints. Monthly Music Hackathon is free of cost, open to all, welcoming to people with all skill levels and disciplines, non-competitive, does not allow advertising or recruiting, and is organized by volunteers who are passionate participants. For more information, please visit the blog, twitter, and facebook pages.