Special Studies in Music
Special Studies in Music MUSIC 680
Popular in Course
Popular in Music
Dr. Zora Franecki
verified elite notetaker
This 4 page Class Notes was uploaded by Dr. Zora Franecki on Tuesday October 27, 2015. The Class Notes belongs to MUSIC 680 at University of Wisconsin - Milwaukee taught by Loren Burns in Fall. Since its upload, it has received 43 views. For similar materials see /class/230261/music-680-university-of-wisconsin-milwaukee in Music at University of Wisconsin - Milwaukee.
Reviews for Special Studies in Music
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 10/27/15
Music 680 Special Topics in Music Interactivity and Improvisation Spring 2007 Class 3 February 7 2007 Pierre Boulez explosunte fixe Jonathan Harvey 0112 Evening the aesthetics of score following and the basic technical aspects more on audio feature detection from Roads Computer Music Tutorial pitch estimationdetection as an inherently noisy phenomenon microvariation vibrato inharmonic instrument tones etc polyphony and musical nonmusical audio context attack transients are typically unstable difficulties recognizing low frequency tones because waveforms are long elapsed time to repeated cycle pitch estimation strategies zero crossing detection or peak interval measurement must discard high frequency information to work correctly can be improved by using a filterbank preprocessor use only the two lowest bands with significant amplitude autocorrelation compare a signal with delayed copies of itself and look for similarities works best for speech limited pitch range reduces computation adaptive filtering tune a comb filter to minimize the filtered input signal results in best fit for relatively harmonic signals spectrum analysis short time Fourier transform produces frequency data from time data seek strong peaks harmonic relationships etc time frequency tradeoff still problematic for low frequency estimation tracking phase vocoder improves on STFT by interpolating inside frequency bands cochlea models understanding of the human auditory system is work in progress rhythm recognition onset detection amplitude thresholding large amplitude increases indicate the onset of a new event ideally combine this strategy with frequency domain analysis is the timbrepitch changing along with the amplitude tempo tracking space onsets into a metrical grid use a decaying history of onsets time windowing to account for uctuations score following allen and dannenberg 1990 and multiple simultaneous representations of location allen and dannenberg tracking musical beats in real time icmc 90 dannenberg and mont reynaud following an improviation in real time icmc 87 amplitude tracking env as a tool to capture RMS amplitude amplitude averaged over very short periods of time takes one parameter length of the time window specified as a number of samples power of two 64 128 256 512 1024 etc reports amplitude expressed in dB where 0 is silence and 100 is the maximum output of the system though prior to the output the amplitude can be much higher if necessary not peak to peak amplitude which describes the maximum and minimum excursions but how do we know where to look for those as a rough rule of thumb add 3dB to the RMS value to get the peak to peak value envelope following as a simple if imperfect way to segment an audio stream look for a new maximum above the current threshold and call it a new event then set the threshold to that maximum and let it gradually decay can be improved by halting the threshold decay for sustained signals and by looking for relative silences as an opportunity to restart or accelerate the decay dbtorms and rmstodb objects for easy units conversions pitch tracking fiddle as a tool to capture a variety of frequency data from an audio stream based upon Fourier analysis transforming a set of time samples into a set of frequency samples uses the tracking phase vocoder to locate specific frequencies inside the frequency bands that are reported by the basic STFT pitch estimation algorithm likelihood of a given frequency peak being the fundamental pitch is determined thus for each peak that s a near multiple of the analyzed frequency multiply together three values the peak amplitude a value expressing how close of a multiple the frequency is a value favoring low multiples over high multiples then sum together the resulting product for each near multiple peak the frequency with the highest sum is the pitch estimate if we re looking for more than one pitch polyphonic estimate then the frequency with the next highest score that isn t a near multiple becomes the next estimated pitch for as many pitches as we ask to estimate takes four default values number of samples per analysis frame defaults to 1024 unless specified again specified as a power of two 64 128 256 512 1024 2048 4096 8192 increasing the number of samples improves pitch resolution decreases time resolution number of polyphonic voices to track defaults to 1 unless specified note that you will get extra outlets if you ask for 2 or 3 voices but don t expect polyphonic pitch tracking to be unproblematic number of frequency peaks to track defaults to 20 unless specified number of frequency peaks to report defaults to 3 unless specified and returns five or more outlets outlet 1 cooked pitch outlet fiddle s best estimate as to pitch and as to changes of pitch outlet 2 report of new events fiddle s best estimate as to the onset of new events outlet 3 quotraw pitch outlet list with continuous estimate of pitch and amplitude for the first polyphonic voice use unpack to pull apart pitch and amplitude into separate values outlet 4 and 5 only if polyphonic voice tracking is specified same as outlet 3 for the second and third polyphonic voices outlet 4 or 5 or 6 continuous estimate of overall amplitude outlet 5 or 6 or 7 data about individual frequency peaks in the overall analyzed spectrum all frequency peak data is reported in list form via the last outlet data is reported as a triple peak number from 1 to the maximum number of peaks to report continuous frequency in Hz for that peak continous amplitude in dB for tha tpeak mtof and ftom objects for easy units conversions prep for next week arraystables data collection storage and representation David Tudor Neural Synthesis n0 8 electronic systems to facilitate and confound improvisation Roger Dannenberg on music representation Music 680 Special Topics in Music Interactivity and Improvisation Spring 2007 Class 11 April 18 2007 EAMC concert Thursday April 19 730 pm including music by Alvin Lucier 2 Pauline Oliveros Kenilworth Open house Friday April 20 5 9 pm and Saturday April 21 12 5 pm last chance for participation Saturday afternoon group improvisation 2 pm setup 230 pm start Karlheinz Stockhausen Karzwellerl as a faulty process of repetition material varies as it is passed across the ensemble due to instrumental difference and because of the variation structure specified by the graphic score the radios as a unique random seed for each performance literally pulling music out of the air Qaartet demonstration block diagram offline pitchtracking online interactivity entirely through a MIDI like protocol which means that the listening diverges from the sound for longer events signal processing techniques dynamic range processing waveshaping as a strategy for limiting compression and expansion env as an average detector limiter abstraction as an example of look ahead compression filtering as an application of delay preview April 25 signal processing techniques using delay more Roads Tutorial reading Pauline Oliveros ear 0n the Road Kaija Saariaho Nymphea faralirl Secret III Olga Neuwirth Instrumental Inseln aas quotBahIamms Fest H