Subject: Re: Automatic transcription of rhythm - advice requested From: "M.F. McKinney" <mckinney(at)alum.mit.edu> Date: Tue, 1 Jul 2003 10:34:34 +0200
Andrew, For tempo and beat tracking, there is a more recent article by Jean Laroche: "Efficient tempo and beat tracking in audio recordings", J. Audio Eng. Soc., V51, N4, April 2003. You could also check out the webpages and articles of Anssi Klapuri: http://www.cs.tut.fi/~klap/iiro/ A difficulty with tempo tracking is that the perceived beat of a piece of music is often ambiguous. This is true even for musicians who are unfamiliar with the notation of a piece to which they are listening. Also, Fabien Gouyon had a recent paper at the AES convention on classifying audio tracks as either duple and triple meter: "Determination of the meter of musical audio signals: Seeking recurrences in beat segment descriptors.", 114th AES Convention, Paper 5811, Amsterdam Netherlands, 22-25 March 2003. Good luck. Martin McKinney ------- "If we knew what we were doing, it wouldn't be called research, would it?" Albert Einstein (1879-1955) ----- Original Message ----- > Date: Mon, 30 Jun 2003 14:25:49 +1000 > From: Andrew Nesbit <alnesbit(at)STUDENTS.CS.MU.OZ.AU> > Subject: Automatic transcription of rhythm - advice requested > > Dear List, > > I wrote several weeks ago concerning my starting out in CASA research, > and I received very helpful advice. I have now narrowed down the > scope of my project, and I am requesting guidance in concentrating my > reading to the most relevant published research. > > We have digital recordings of Indigenous Australian music. It > contains three instruments: yidaki (didjeridu), vocals and clapsticks. > One of each instrument is present in the mix. What makes this project > interesting is that the music contains polyrhythms, changes of metre > and changes of tempo. My task is to develop a system (in Python) to > automatically transcribe the rhythms in this music. Transcription of > pitch and other musical characteristics would be a bonus, but it is > not the main goal. The system is not a general-purpose CASA tool, but > rather, tailored towards a specific application. > > Does anybody have any suggestions for published research which would > be helpful for me to get insight into my project? > > (My overall plan, so far, is to separate out each of the instruments > using various instrument models. Next, perform detection of note > onsets and other salient rhythmic features in each of the separated > parts. Then, using an (as-yet undetermined) "grid-laying" algorithm, > which operates within a given part and also across different parts, to > find a best-fit combination of metres and tempi which satisfy the > rhythmic characteristics of the original mix.) > > Finally, is this the best list to be asking such questions? If there > is a more appropriate forum, could somebody please direct me to it? > > Thankyou in advance. > > Andrew.