CSDT Work Monday 6/20-Sunday 6/26

Monday 6/20 9:15am-12:15pm:

• From last Friday, calculated fractal dimension was not matching input fractal dimension
• This is because analysis was being limited to a certain frequency range within the whole power spectrum
• Analysis of whole power spectrum confirms calculated fractal dimension equals input fractal dimension
• Produced graphs of various known power distributions

Tuesday 6/21 9:30am-12:45pm:

• Measured processing time for analysis script for audio samples with various numbers of datapoints
• Found processing time increased quadratically with number of datapoints
• Nearly all time spent on calculating instantaneous pitch
• Tested ways of decreasing processing time
• Worked on computing instantaneous pitch in parts and concatenating

Wednesday 6/22 10:15am-11:00am, 12:30pm-3:45pm:

• Continued work on computing instantaneous pitch piecewise
• Attended weekly team meeting
• Last week's goals accomplished up to loading the entire classical piece
• Need to edit script to allow loading audio piecewise
• Goals for next week:
• Today, Thursday:  Finish editing script to process pitch of long audio files by breaking into parts
• Friday: Minimize processing time for algorithm; analyze 7 minute classical piece
• Monday: Repeat some of the analysis done in Professor Eglash's paper
• Analyze several samples of poetry reading, prose reading, and a Capella music
• Chart fractal dimension for each sample
• Tuesday: Analyze rhythm produced using CSDT rhythm wheels
• Wednesday (6/29): Prepare for meeting

Thursday 6/23, 10:00am-3:00pm:

• Compiled photos, arc coordinate printouts, and screenshots from teaching Anishinaabe Arcs at Albany Boys and Girls club into zip archive
• Finished script for analyzing wave files piecewise
• Need to read the file piecewise
• Script run time based amount of data Matlab has in workspace (not necessarily in scope)
• Corrected fractal dimension calculation

Friday 6/24, 12:00pm-12:30pm, 3:30pm-5:45pm

• Optimized script to find size of pieces to use to minimize run time
• Tested script on known fractal dimension-audio and short Mozart audio clip

CSDT Work Monday 6/13 - Sunday 6/19

Tuesday 6/14 9:15am-12:00pm, 1:45pm-3:30pm, 8:00-8:30pm:

• Continued comparison of methods for computing instantaneous pitch of music file
• Took notes on David Gerhart report (This is the report.)  
• Compared time- and frequency- domain approaches, and various specific algorithms 

Wednesday 6/15: 1:15-2:45pm, 8:00-9:15pm:

• Read through report by UPF Music Technology group on melody extraction contest
• Found toolbox for audio analysis
• Includes pitch and salience extraction, segmentation, and many other features
• Will try implementing in MATLAB

Thursday 6/16: 8:15am-1:00pm, 2:30pm-3:45pm:

• Used audio analysis toolbox to compute pitch waveform for several audio samples using autocorrelation
• Experimenting with methods of smoothing or possibly quantizing pitch waveform

Friday 6/17: 12:30pm-2:00pm:

• Attempted to verify MATLAB script I am using to compute power series
• Ran power series analysis for noise generated with MATLAB algorithm 
• Tested 1/f^alpha for 15 values of alpha from 0 to 3.5
• MATLAB script returned alpha values that were consistently 
• 17% off for 1 sec samples
• 31% off for 2 sec samples
• about 40% off for 3 sec samples
• Will test another noise generator and possibly another method of computing power series

CSDT Work Mon 6/6 - Sun 6/12

Monday 6/6 11:00am-12:15pm, 3:30-5:00pm:

• Completed Matlab script to plot best fit line with Power Series of Amplitude and display slope
• Helped teach Arcs applet to Boys and Girls Club in Albany
• Students constructed circular wigwams
• Located points on life-sized grid (outside)
• Attached saplings to grid using ties through holes drilled in saplings
• Students seemed more comfortable with finding points on graph
• Knew if given (-4,3), would have to move 4 in one direction and 3 in another from (0,0)
• Often did not distinguish between x and y and positive and negative, especially for 3rd and 4th grade
• Students enjoyed building wigwam
• Offered an impromptu lesson on finding area using life-sized grid

Tuesday 6/7 10:10-10:40am:

• Began researching methods for finding instantaneous pitch of audio signal
• Aim is to find power series of pitch
• Analyze various musical genres and other audio building on Voss and Clark work (Nature, Vol 258 Nov. 1975, p. 317-318)

Wednesday 6/8 10:20-10:50am, 1:30-3:15pm:

• Continued researching methods for finding instantaneous pitch
• Did not find any algorithms for using solely zero-crossing 
• Began testing algorithm using "subharmonic to harmonic ratio" (geared towards speech)
• Attended weekly planning/review meeting
• Goals for next week:
• Tues 6/14: Research pitch-detection algorithms
• Wed: Test algorithms and compare performances; prepare for meeting
• Thurs: Help compile report on teaching Anishinaabe Arcs at Albany Boys and Girls Club
• Fri: Successfully compute instantaneous pitch of short audio clip using chosen method
• Mon: Load longer (~7 min) classical composition piecewise into Matlab; 
• Compute amplitude power series
• Compute instantaneous pitch
• Tues: Catch up; prepare for meeting
• Next goal: compute power series of instantaneous pitch

CSDT Work Monday, May 30 - Sunday, June 5

Monday, 5/30 10:30am-11:45am, 2:45-3:45pm:

          Found a Matlab script to run Fast Fourier Transform (FFT) and compute power series http://faculty.olin.edu/bstorey/Notes/Fourier.pdffunction)

• Adapted script to operate on input from .wav audio file

• Ran analysis on Bach pieces and white noise sample

• Did not find expected slopes for Power Series 

• 
  

• Tuesday, 5/31 12:00-2:45pm: 

• Bach pieces were recorded with instruments that did not vary volume (amplitude) with time.

• Ran power series analysis on Mozart piece played with violin, sample of 

• white noise (alpha = 0), and 1/f noise (alpha = 1)

• Did find expected correlation; Mozart slope appears close to alpha = 1:

Power Series analysis of Mozart Piece and Alpha = 0 and Alpha = 1 Noise

Wednesday 6/1 10:25-11:25am, 2:00pm-5:00pm

•      Edited Matlab script to select only the data in frequency range of interest (~1/(time of piece) to 0.5Hz))

•      Worked on setting script to compute slope of best fit line of LogLog plot

•      Attended meeting with design group; Goal: Run power series analysis on pitch 

•        sequence by this weekend

•      Helped transfer screenshots and print-outs of coordinates for Albany BGC Anishinaabe Arcs project

Friday 6/3 1:15-2:15pm, 3:30-5:30pm

•      Printed out Anishinaabe Arcs Projects

•      Taught Lesson #3 of Arcs Project at Albany BGC

•           Had students 1) Label graph paper with coordinate axis

•                        2) Plot point from their arc printouts

•                        3) Attach printouts to styrofoam and add bendable rods to 

•                           create their wigmwams

•      Students had far more difficulty with this than either of two previous lessons

•      3rd and 4th grade students were not familiar with using coordinate axis; they

       labled the axis, but not at even intervals, and they needed individual instruction

•      Probably 3 or 4 of about 16 students were comfortable plotting points without any 

       help by end of the session; some other students made significant progress

•      Students who learned helped their classmates

•      Perhaps target age group for activity could be 5th or 6th grade 

Saturday 6/4 2:00-2:45pm

     Worked on getting best fit line to logrhythmic plot of Power Series
     Difficulty adapting linear best-fit to logrhythmic plot