Project Proposal – “18 is basically 20”

General support to our idea: here

Metrics that define success:

 

• Pitch similarity percentage until annoyance

 

• • • Absolute vs relative vs perfect pitch. We care about absolute pitch
    • Like most human traits, AP is not an all-or-none ability, but rather, exists along a continuum 10, 17, 20, 21. Self-identified AP possessors score well above chance (which would be 1 out of 12, or 8.3%) on AP tests, typically scoring between 50 and 100% correct [19], and even musicians not claiming AP score up to 40% [18]. Here
    • AP possessors incorrectly identify tones that are 6% apart here (best possible case to meet AKA the hardest) – upper bound of accuracy
    • The response accuracy of melody comparison is shown separately in Figure 2 for the AP group and the non‐AP group. The chance level is 50%. In the C major (non-transposed) context, in which the two melodies could be compared at the same pitch level, both the AP and the non‐AP groups gave the highest level of performance; in contrast, in the E– and F# context, in which the comparison melody was transposed to different pitch level from the standard melody, both groups performed markedly worse. Notably, the AP group performed more poorly than the non‐AP group. Here (should our percentage of annoyance error depend on the pitch of the song?)
      • Non AP scores 40-60
      • AP scores 80-100
      • Avg = 70?
    • 1-5 people/10,000 have AP here => don’t worry about AP

 

• Pitch will remain same with 25 cents (¼ semitone) marginal error

 

• • Will we have a relative pitch problem as songs don’t stay at one tone throughout?

 

• Percentage of difference between pace and tempo until annoyance (pulsing)

 

• • • Helpful pace/tempo matching Here
    • Helpful pace/tempo matching Here 
    • Runner’s side bpm measure here
    • 120-140 bpm is normal
    • Over 150 bpm probably too fast & will be distracting – we will ensure to stay under 150 bpm.
    • Range should be (inclusive of walking) 90-153 bpm

 

• Room for error in pace detection (stride deviation)

 

• • • https://www.researchgate.net/publication/321413174_Sampling_Frequency_for_Step_Detection_Based_on_Smartphone_Accelerometry
      • Based on different algorithms and different sampling frequencies, the error rate for step detection was close to 4% for the pocket
    • https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4634483/
      • S-Health – 94.4% accuracy
      • i-Health – 82.3% accuracy

 

• How long between change in pace and change in tempo of currently playing song

• https://www.sciencedirect.com/science/article/pii/S0966636210002717

• Standard change in pace over run seems to be around + or – 5%

• How to connect sensor on shoe to phone to share data

 

• • • New pedometers have integration with phones using bluetooth and different apps

 

• Which sensors to use / how to use sensors 

 

• • • Best sensor would be a pedometer that you can attach to your shoe that connects to your phone using bluetooth
    • We can use this to track how accurate the step count is verse a phone/smart watch

 

• Real-time feedback – how often?

 

• • • Instead of time, let’s use # of footsteps
    • Since this gives a better relation to bpm than time does
    • Remeasure & calibrate every 20 steps? (conjecture based on my running experience. We should test amongst us & friends / random gym people for design proposal (mention this in project proposal)) 

 

• calibration? 

• Song choice algorithm

• https://www.independent.co.uk/life-style/gadgets-and-tech/news/why-random-shuffle-feels-far-from-random-10066621.html

• Spotify was forced to not make the song choice truly random

• https://digest.bps.org.uk/2018/03/09/psychologists-have-explored-why-we-sometimes-like-listening-to-the-same-song-on-repeat/

 

 

Shopping List:

• Bluetooth wireless headphone(s)
• Sensors
• Extra smartwatch?
• Android phone (ask Sullivan for 18-551 extras)

 

* scratch ML algo for tempo detection

* this app end goal is to be linked to spotify → metadata of songs

* long distance runners use this to help maintain