HearingProjectMinutes

From Norsemathology
Jump to navigation Jump to search

HearWell

November 29, 2022: 8:00

Biplov

  • Demo of testing procedure
  • Follow-up on our call with Tommy?

Yoseph

  • LoCHAid update?
  • Have we got a poster started?

Andy

  • We need to finalize the travel plans for the JMM. Flights? I'll take care of the registrations, but I'm hoping to get this all done today.
  • We need some names for our hardware person for the project. Have we got any?
  • Some action items from Tommy:
    • He is willing to look at a file (saved to Excel) and see if our ftp library is behaving
    • Suggested we get a DSP, to handle the sound input (and output?)

Steve

  • Python to Yoseph?

November 21, 2022: 8:00

Biplov

  • Block diagram?
  • Demo of testing procedure

Yoseph

  • LoCHAid update?

Andy

  • Tommy's list:
    1. Get a timeline on when it needs to be spent -- End of February, 2023
    2. Make a generalized block diagram of what’s needed to make a first (or second?) prototype happen
      1. This might even include doing things with high quality implements on the PC side before jumping to microcontroller (a USB mic, 16-bit DAQ, LabVIEW or Matlab for processing, etc.)
      2. At a very high-level, what kinds of sensors (mics), actuators (speakers), and processing power plus measurement quality we’ll need (micros are very different: some analog inputs are only 10-bit, some are 12-bit (4 times the resolution), etc.).
    3. Soldering tools I can suggest (I use HAKKO FX888D-23BY or similar), but we also have the funds to do PCBA, where the fab house can put your parts down and you get fully (or almost fully) populated boards.
    4. I have access to a 25-seat license of Altium Designer, and can provide access to anyone with an EDU email address. Tutorials on YouTube can get anyone started and UofL has an Altium Workspace for tracking design iterations (actually uses GIT).
    5. Create a GIT repository for code development (my username is tjrous02).

Steve

  • Python to Yoseph?

November 15, 2022: 8:00

Biplov

  • We need to transition over to Yoseph; maybe you two want to meet separately and inventory, and so on. Let's get a plan together for this.

Yoseph

  • LoCHAid update? Did you hear back from the folks at the lab?

    The comments at the end of the video are enlightening; no replies from the lab, unfortunately.

    There is a purported parts list; don't know if it's the same stuff Biplov ordered.

Andy

  • Tommy's list: "Before we blow any of that $$$, let’s:
    1. Get a timeline on when it needs to be spent -- End of February, 2023
    2. Make a generalized block diagram of what’s needed to make a first (or second?) prototype happen
      1. This might even include doing things with high quality implements on the PC side before jumping to microcontroller (a USB mic, 16-bit DAQ, LabVIEW or Matlab for processing, etc.)
      2. At a very high-level, what kinds of sensors (mics), actuators (speakers), and processing power plus measurement quality we’ll need (micros are very different: some analog inputs are only 10-bit, some are 12-bit (4 times the resolution), etc.).
    3. Soldering tools I can suggest (I use HAKKO FX888D-23BY or similar), but we also have the funds to do PCBA, where the fab house can put your parts down and you get fully (or almost fully) populated boards.
    4. I have access to a 25-seat license of Altium Designer, and can provide access to anyone with an EDU email address. Tutorials on YouTube can get anyone started and UofL has an Altium Workspace for tracking design iterations (actually uses GIT).
    5. Create a GIT repository for code development (my username is tjrous02).
  • For some with ADHD, the low rumble of brown noise quiets the brain: Listening to brown noise could have cognitive benefits for people with ADHD, but experts caution the evidence is still limited

    Includes samples of brown, white, and pink noise. We're ultimately going to have to deal with noise, and this article

Steve

  • Any luck finding the adaptive poster?

November 8, 2022: 8:00

Biplov

  • File with FFT data?
  • The I-Corps funding remaining is $1,355 -- we can spend it on prototyping, so why don't you propose how we spend it, in conjunction with Yoseph.
    • Be specific. We want prices and URLs -- but I think that we have to let them order for us (still have an email out about that).
    • How do we use it to get out of the gate quickly, with as much flexibility as we can?

Yoseph

Andy

  • Finishing Biplov's Capstone (Due at the end of the semester -- just a few weeks!)
  • Preparing for the JMMs
    • Logistics
    • Yoseph has a conflict with a training for residence hall work. Will see if he can wiggle free, and I may need to send a letter.
  • Testing: the Adaptive scheme.

    This from an earlier time (in July):

    Adaptive testing: Let's talk about how this might work. I'm basing this example on "adaptive quadrature" -- the stopping criterion is determined by when you might imagine that you've got enough of a sample of a function to produce a pretty good estimate of the integral of the profile between min and max, using Simpson's rule.

Steve

  • Metrics?

November 1, 2022: 2:00

Biplov

Yoseph

  • LoCHAid? Yes! Has achieved contact.

Andy

There are three things, in particular, on my mind:

  • Finishing Biplov's Capstone
    • Due at the end of the semester -- just a few weeks!
    • Needs to have some sort of device -- a prototype. Let's focus on the testing, and put getting the display working onto the back burner.
    • So we can generate frequencies, but we can't yet pick off particular ones to display -- Biplov will try using Bandpass filtering, to see if that works with the display -- but that may be for a later date; not for the capstone.
  • Preparing for the JMMs
    • Our focus is on the adaptive scheme for testing
    • Logistics
    • Yoseph has a conflict with a training for residence hall work. Will see if he can wiggle free, and I may need to send a letter.
  • Transitioning to KYNETIC
    • I've reached out to Partners In Health (PIH) --
    • We need to get IRB approval squared away (eventually)
    • We need to find the technical team member, who can take over the hardware side.

Steve

  • Class conflict, so couldn't make it.

October 18, 2022: 8:00

Biplov

Yoseph

  • Any issues with the Joint Math Meetings? We'll need to get the travel squared away relatively soon; we can run the registration through the department, I imagine.
  • LoCHAid?

Andy

Steve

October 11, 2022: 8:00

Biplov

Yoseph

Andy

Steve

October 4, 2022: 8:00

Biplov

Yoseph

Andy

Steve

September 27, 2022: 8:00

Biplov

Yoseph

Andy

  • WHO's Preferred profile for hearing-aid technology suitable for low- and middle-income countries
    • Include Table 1. Modified WHO performance requirements (WHO, 2004)
    • There there are "Essential" features of hearing aids
  • Neosensory's Clarify uses vibration to help on "retrain the brain" (they actually use this expression...)
  • LoCHAid: An ultra-low-cost hearing aid for age-related hearing loss

    • Science article
    • Blueprints!
    • $0.98 per unit, if you buy 10,000...:)
    • Its simplicity has a few drawbacks, however. The device can't be fine-tuned for individual needs or adapted to treat other hearing issues. And even though it's waterproof and shock resistant, the scientists anticipate that LoCHAid's parts will wear out after about a year and a half. Its bulky size might deter some users, Bhamla says, although a smaller version is in the works.
    • The next step is getting people to use the devices. In low- and middle-income countries, only 3% of people with age-related hearing loss wear hearing aids, and in countries like the United States, the adoption rate hovers at about 20%, according to Bhamla. Aside from concerns about cost, Bhamla says, "a lot of people don't realize they have hearing loss … and then there's the social stigma of wearing the aid."
    • Georgia Tech's page about the device
    • I don't find much recent info about it (this is mostly from 2020).
    • Biplov: would it be possible for you to order the parts, and then build it? Maybe two of each, and both you guys build it.

  • KYNETIC update
    • The grant proposal is in.
    • We had to scale back the team a bit, due to budgetary reality; at the moment, it's got funding for just two students -- a "lead" and an "assistant".
    • We do have funding for Tommy Roussel built in, so we'll have him to lean on for technical stuff if the funding comes through.
    • Yoseph:
      • Per the future, if we're funded: there's a "student leader" position in the budget, for 13 hours per week, $15/hour. Can you swing that for the spring?
      • There's also a "student assistant" position: do either of you have any candidates in mind? Seems like we'd need some hardware help, in particular.
    • Selected references for the grant proposal

Steve

September 20, 2022: 8:00

Biplov

Caption

  • Questions for Tommy?
  • Graphic of the harness for the grant proposal?
  • Here's a graphic that may be useful for the poster
  • All set for Boston! Great that you got that moved around. Now the work begins, of course.

    Why don't you start a poster, and we'll begin to fill in the various sections. One thing we need to do is avoid proprietary disclosures (and I fear, by the way, that our abstract may already gives away more than we should - que sera, sera...).

Yoseph

  • Aren't you the World's Expert on What's Available over-the-counter? Would you please give some thought to these sections of the Kynetic proposal (and can you give me your feedback by Thursday's LaunchIt meeting)?
    1. Market Size: Define the total and addressable market size, and target price of the technology. Support your market size and descriptions with evidence about current technologies or approaches to address this indication. Define a specific patient segment of those suffering from the specific targeted disease. What are the market population trends and projections?
    2. Competitive Landscape: Define the competition mix (companies, products, substitutes and shifting landscape) for the technology. Think particularly about how the disease will be treated when the technology/product gets to market. How is the landscape shifting or projected to shift?

Andy

  • Good recovery, Biplov, on the JMM poster abstract. How shall we prepare? How does this dovetail with your capstone?

  • KYNETIC update
    • I'm awaiting revisions from the KYNETIC folks.
    • We had to scale back the team a bit, due to budgetary reality.
    • Yoseph:
      • I'm hoping that you can kick in on the competitive landscape.
      • Per the future, if we're funded: there's a "student leader" position in the budget, for 15 hours per week, $15/hour. Can you swing that for the spring?
      • There's also a "student assistant" position: do either of you have any candidates in mind? Seems like we'd need some hardware help.

  • LaunchIt Thursday.

Steve

September 13, 2022: 8:00

Biplov

  • Update on Arduino issues -- even if we may want to migrate away from Arduino....
  • Purchases?
  • Light show?

Yoseph

  • Ready for LaunchIt?
  • Can you be the World's Expert on What's Available over-the-counter (by Thursday?):) See the New York Times article below....

Andy

  • JMM Poster Session Abstract: today's the deadline! Reminder that I sent the following:
    Hi Guys,
    
        I chased down the info on the poster session. See if the link below provides all you need..... I created an account, and got to the point at which I could submit an abstract for the undergraduate student poster session.
    
        We'll want to wait to actually do that, of course, until we've agreed on an abstract. 
    Deadline for New Submissions: Tuesday, September 13, 2022
    
    Andy
    

    What shall we offer to share? Let's do it under the category of "65 Numerical analysis".

    • Working title: The wrong tool in the right place: better hearing through adaptive quadrature and curvature
    • using adaptive quadrature (and maybe adaptive curvature) to choose which frequencies get tested
    • additional focus on the DFT (the right tool in the right place), and putting the weights to the spectrum for better hearing
    • A fun quote: George Thorogood and the Destroyers (Who Do You Love?): "You should have heard just what I seen..."

    It would be nice if we get the website up and working, so that we could record someone's voice, and then playback their voice with an HLP (Hearing Loss Profile) applied to it; they could hear just what I need done to their voice to hear them "properly". (It sounds improper to the speaker!)

  • KYNETIC update
    • No word yet on a meeting with Tommy -- If I get some questions from Biplov, I'll try to get Tommy for Tuesday next.
    • Steve: can Taraneh do a little artwork for us? (Or is there someone else?) -- Biplov will try to get a 3D animation we can use, and maybe Taraneh can doctor it up; add some hair (or not)....
    • New York Times pointed me yesterday to this site: The best over-the-counter hearing aids.
      1. Adjustability is key: Situational sound modes let you change what sounds get amplified in different environments, which means better clarity.
      2. Give it time: Your brain needs to adapt to the sound of hearing aids. A generous return policy gives you more time to acclimate.
      3. Best earbud-style hearing aids for iOS users: Jabra Enhance Plus: The Jabra Enhance Plus set is one of the first hearing aids released with the goal of meeting FDA over-the-counter regulations. (about 800 bucks)
      4. Tweak Enhance: A simple, affordable PSAP
    • A fun quote:
      • George Thorogood (Who Do You Love?): "You should have heard just what I seen..."
  • LaunchIt update -- Thursdays, 4:00-5:30; we have homework! I'll take the lead on that....

Steve

  • Python issues?

September 6, 2022: 8:00

Yoseph

Biplov

Andy

  • KYNETIC
    • Folks: Introducing the KYNETIC team, of Jamie, Jessica, Sarah, and Megan (of KCV); also Tommy, from U of L.
    • grant draft
      • application
      • budget: proposing team of five (Andy, Steve, Yoseph, two others); plus consultants

Steve

August 30, 2022: 8:00

Welcome Yoseph!

Biplov

  • What's on the agenda for purchases?
    • Noise cancelling headphones?
    • The LED light experiments? (This one will be especially important for creating a light-frequency version of the sonic signal.)
    • Thoughts about making this smaller and cooler. Building a shipping container to do the testing, with a smaller unit inside to serve as the hearing device.

Andy

  • Student poster presentation at the Joint meetings: looks like they're on (based on this website)
  • dft as matrix multiplication
    1. Second section from the bottom: sampling using six data points at a time....
  • Yoseph's role:
    1. Support Biplov in hardware and software development
    2. potentially provide continuity, should the grant be funded (depending on interest)
    3. help prepare the grant application

Steve

  • Python issus?


August 19, 2022: 3:00 -- in person

We met outside of Steve's office, to discuss progress so far, and future directions. The exciting news from Andy's perspective was that we have been invited to make a submission of a "full application". We have discovered that we're 1 of 3 projects for two slots; so the odds are good, but we have to make this look exciting to the KYNETIC folks.

Biplov

  • Biplov is moving away from the Arduino
  • Andy encouraged Biplov to spend our budget for materials; include light displays (this is one of our important special tricks)
  • Andy also suggested that we consider making a flashy version of the testing device for the "shark tank" meeting; this will meeting will serve as Biplov's presentation requirement for the summer research project. (What other reporting requirements are there?)
  • Question about student poster presentation at the Joint meetings.

Andy

  • Will look into the dft as matrix multiplication
  • Will pursue adding Yoseph to the project, to provide Biplov some help. Yoseph will also potentially provide continuity, should the grant be funded.
  • Andy seems to recall discussing using the box in which the device is sent to beocme the hosing for the tester; then, when that's done, there a smaller box inside that one could use as the hearing aid.

Steve

  • Continues to pursue the python tester.

August 8, 2022: 10:00

Biplov

Andy

  • On-line hearing tests -- everyone has one! I tried the Phonak's tester, and here are my thoughts
    • It was called the three-minute tester
    • They tested only two frequencies for each ear: 2000 and 4000 for the right, 1000 and 6000 for the left. (Why?)
    • The tone beeps -- pulsates -- which makes a certain amount of sense. The steadiness of the pulsation helps one determine that one is hearing the tone, rather than background noise.
    • You have to provide your email, and approve that phonak will contact you with product info, in order to get your results
    • Yet their report suggests that they have predictions for 6 tones, both right and left! Quite an extrapolation....

  • How shall we use the profile? In my Bluegrass file, I multiplied each frequency (band) by the appropriate profile value, and then, in the end, had to adjust the total volume to make the manipulated file (sum of products of band with amplitude for that band) sound like it was the same loudness as that of the Phonak-filtered music (by a factor of 10, roughly).

  • Mathematica
    • The help for ShortTimeFourier has a really interesting application of audio manipulation, including things like "Define a nonlinear function to squash low-amplitude components"

      I think that we can use this to do some of the "profile adjustment". Again, it's not on the fly. We need to be doing this on the fly, with the Arduino.

    • Creating "non-linear squashers" may be that "micro-bandpass-filtering" that we need....

  • Comment on the adaptive quadrature: since the intervals along the x-axis (frequencies) are growing exponentially, I "grew" the epsilon exponentially, too: I had a starting epsilon (for the first octave), and then doubled it for each successive octave: -- something like that.

Steve

August 5, 2022: 9:00

Biplov

Progress on adaptive testing implementation?

Andy

Steve

August 2, 2022: 9:00

Biplov

Progress on adaptive testing implementation?

Andy

  • Mathematica
    • Here's a standard hearing aid, demonstrated to a bit of Bluegrass (in Mathematica). It works off Bandpass filters.
    • I was going to use hash tables on my adaptive testing code, but they weren't introduced until 2020 -- Mathematica 12.1 -- and I'm stuck back on 12.0. I did write a version of it using them, and Steve if you'd test it that would be swell. This code might work....:)
  • FYI: here's the KYNETIC grant proposal, which includes a fair bit of background and planning (if all goes well).

Steve

July 27, 2022: 9:00

Discussion of adaptive scheme. Steve went over the basic idea, and, although the curvature method might be better for plotting most functions in calculus, he feel like a Simpson's rule-based scheme would be simpler, and better for this project. One question is how to set "the epsilon" (stopping criteria); Andy suggested that is should be relatively straightforward, given that the shapes of audiograms are probably fairly uniform. Surely we won't see too many vertical asymptotes?

Andy mentioned that the audiogram is, in some sense, a cochlear response map -- it's a picture of your cochlea's ability to produce sound (or of the Cochlear nerve's ability to transmit that sound to the rest of the brain):

  1. inside the cochlea
  2. from Gray's Anatomy (1918)

Andy hoped that the device will be easily re-programmable. (Biplov suggested just retesting, but both Andy and Steve have noted the tedium of testing). So if one could do part of the testing today, and continue tomorrow (and update/refine as needed) that would be good.

July 25, 2022: 9:00

Biplov

  • Reports on
    1. Shokz headphones -- bone conduction. Need a transducer.
    2. Can we try that arduino light display experiment?

Andy

  • KYNETIC grant -- where are we?
    • Proposal so far
    • A few things that might help me get this grant proposal done....
      1. How will the device be powered in a rechargeable fashion?
      2. Can we manufacture the wearable device, e.g. glasses? Easier to use available ear buds, I'd say.
      3. How much might this ultimately cost?
      4. Consumer Reports has good information about "competitors" in a 2021 report: "With the right fit and adjustment, we found that the higher-end models [ael: of Personal Sound Amplification Products (PSAPs) -- approx $200-350] can help those with mild to moderate hearing loss, especially when watching TV. Some adjustable models can even have the same functionality as an entry-level hearing aid. But beware the penny-saver PSAPs: The cheaper options didn't measure up, and more important, our expert found that they could potentially damage hearing if used long-term - by over-amplifying some loud sounds, such as a fire engine siren."
      5. Biplov, they may only want to see the names of those who will ultimately work on this project, so if you're not going to work on it, your name won't appear; Steve, I'll need some information about you for the grant proposal -- there's evidently a form that you'll need to fill out.
    • Megan Aanstoos may join us at 9:00 for more info.
    • One of the advantages of going for this grant is that I've been forced to do some background research:
      1. Frequency Lowering Fitting Assistants, including
        1. How to Use Probe Microphone Measures with Frequency-Lowering Hearing Aids includes a Glossary of important terms.
        2. s-sh Confusion Test Audio Files, downloadable (here's another that Alexander mentions elsewhere).
      2. 20Q: Frequency Lowering - The Whole Shebang: Scollie: ...one reader questioned if whether a transposer system should be used rather than frequency compression because it will maintain harmonic relationships in music? This is an interesting question that could use more exploration. Because transposers work on octaves, they are described as maintaining harmonic structure, but they may also have other effects on the music signal, such as filtering, that may be benign at mild settings but important at strong settings. I’ve had some patients enjoy music more with frequency compression than without, but have not had the opportunity for a head to head comparison across frequency lowering types for music specifically – remember that these processors were often developed with speech audibility in mind, just like cochlear implants were. We need to find what works for individuals, and this will depend not only on the technology but also on the specific settings used.

        Transposition strategies generally move information to much lower frequencies than nonlinear frequency compression strategies. Even though transposition strategies have been designed to maintain harmonic spacing, because they mix the newly lowered signal with the original low frequency content they might create a peak in spectrum that either masks or mimics a low frequency formant.

      3. 20Q: Frequency Lowering Ten Years Later - New Technology Innovations: "The key differentiators between these earlier methods and the methods now implemented by the Big Six are that modern methods are all digital and they only lower a portion of the speech spectrum instead of the entire range.... The real advantages are that the low and mid frequencies that are already audible without lowering do not need to be altered, or at least altered as much. This helps preserve overall sound quality and reduces the risk of negatively affecting the perception of those speech sounds whose primary cues are in this frequency range, especially vowels (e.g., Alexander, 2016a)."

        "...let’s start from the top of Table 1. The original Audibility Extender and the newer version found in Widex hearing aids uses frequency transposition. Audibility Extender continually searches for the most intense peak in the source region (the frequency range subject to lowering) and then copies (transposes) it down to the destination region (the frequency range where the newly-lowered information is moved to) by a linear factor of 1/2 (one octave). The peak is band-pass filtered so that it only spans one octave after lowering. One advantage of this filtering is that the lowered information does not need to be compressed. This fact along with the linear downward shift helps to ensure that the harmonics of the lowered source align with the harmonics of the signal already present in the destination region."

        "Perhaps, we are seeing a trend toward adaptive frequency lowering because the processing power of the chips in the hearing aids have advanced so much that signal processing artifacts, throughput delay, current drain, etc., are not as much a concern for frequency-lowering algorithms as they once were (see Alexander 2016b for a discussion on these topics)."

        ael: note Figure 2, which is the input to the Starkey HA, which is a primative linear spline:
        "Figure 2. Audiogram entered into the Starkey Inspire programming software that was used to evaluate the effects of Spectral iQ and changes in the Bandwidth setting as shown in Figure 3. The allowable destination region is governed by the audiometric frequency that starts the audiometric slope (1500 Hz in this example) and the lowest frequency where thresholds are ≥ 70 dB HL (4000 Hz in this example)."

        "We know that most of the speech information transmitted in the high frequencies is in the form of frication (the noisy turbulence produced by the narrowing of the structures in the oral cavity). We also know that local and global spectral relationships between phonemes in the fricative class (e.g., peak frequency, center of gravity, spectral shape such as flat, sloping upwards/downwards, etc.) bandwidth, and others) and the speech sounds that precede and follow them play an important role in speech perception (e.g., Kent & Read, 2002). However, we do not have a good knowledge base for how individuals interpret frequency-shifted cues or for what is the best way to transmit them. Obviously, many of the local (frequency specific) spectral cues are distorted after any frequency lowering of any kind. At this point, it is an open-ended question as to how frequency lowering can provide benefit."

    • An Update on Modified Verification Approaches for Frequency Lowering Devices. Danielle Glista, PhD, Marianne Hawkins, MCISc, Susan Scollie, PhD

  • Mathematica
    • Here's an example of the shifting, to a bit of Bluegrass (in Mathematica). It works off Bandpass filters.

      I use AudioPitchShift in this example, which shifts portions of the spectrum of a sound up or down by octaves. Can this be done relatively quickly, on the fly, in the Arduino?

    • Adaptive testing: Let's talk about how this might work. I'm basing this example on "adaptive quadrature" -- the stopping criterion is determined by when you might imagine that you've got enough of a sample of a function to produce a pretty good estimate of the integral of the profile between min and max, using Simpson's rule.

      Each time the function is called, we'd be taking a sample. This example code is wasteful, in that it recalculates some values; it would be better to create a hash table to store values, because we don't want to resample at a frequency. Once and done!

      That's sort of an odd criterion, but makes a certain amount of sense: we are asserting that we don't believe that the profile function is going to go through any more gyrations. Steve's done similar things for plotting a function, so will have used a different stopping criterion. That's an interesting mathematical (as well as biological) question: what's the best criterion in this case?

  • Arduino: I just did a Google search for bandpass filters for the Arduino, and I turned up some interesting stuff:
  • More thoughts on cheap "reading glasses" version of a hearing aid, based on the average profile?

    This reminds me now of the "Personal Sound Amplification Products (PSAPs)" mentioned by Consumer Reports. It seems to me that it depends on whether we can create programmable bandpass filters for the Arduino (or the like).

Steve

  • Adaptive code: moving to Python?

July 19, 2022: 9:00

General

  • Steve is in Japan; back around 7/21

Biplov

  • Reports on
    1. Transform x=0-1023 into N octaves, from 250:

      Note that since your stepper is based on powers of 2, x runs over a power of 2 () steps. Maybe a better way to represent the function is this:

      What we've done is create a composition of two functions: your linear function has become an input to an exponential function (base 2), to create equal treatment of octaves (rather than equal treatment of linear stretches of frequencies).

    2. multiplexer for switching between ears

Andy

  • KYNETIC grant -- where are we? Can we have a look at that NKU disclosure form?

  • Shokz headphones -- bone conduction. Need a transducer.

  • Can we try that arduino light display experiment?

  • Mathematica
    • It occurred to me that Mathematica may be able to record sounds; it can! So we can create some audio snippets, with high frequencies, and then replay them with the shift.... Here's an example
    • AudioPitchShift: This is a great built-in which does this across the spectrum of a sound. So this can be done in theory, relatively quickly.
    • The latest: I've been working on taking a sound (as a simple example [1] here's a whistle), breaking it down using BandpassFilter, and the putting it together with emphasis determined by the audiogram.
    • Loudness: Mathematica seems to use RMS Amplitude, although there are two others that actually include the word "Loudness" in their names.
  • "Linear predictive coding (LPC) is a method used mostly in audio signal processing and speech processing for representing the spectral envelope of a digital signal of speech in compressed form, using the information of a linear predictive model... LPC is the most widely used method in speech coding and speech synthesis. It is a powerful speech analysis technique, and a useful method for encoding good quality speech at a low bit rate."

    Overview

    LPC starts with the assumption that a speech signal is produced by a buzzer at the end of a tube (for voiced sounds), with occasional added hissing and popping sounds (for voiceless sounds such as sibilants and plosives). Although apparently crude, this Source–filter model is actually a close approximation of the reality of speech production. The glottis (the space between the vocal folds) produces the buzz, which is characterized by its intensity (loudness) and frequency (pitch). The vocal tract (the throat and mouth) forms the tube, which is characterized by its resonances; these resonances give rise to formants, or enhanced frequency bands in the sound produced. Hisses and pops are generated by the action of the tongue, lips and throat during sibilants and plosives.

    LPC analyzes the speech signal by estimating the formants, removing their effects from the speech signal, and estimating the intensity and frequency of the remaining buzz. The process of removing the formants is called inverse filtering, and the remaining signal after the subtraction of the filtered modeled signal is called the residue.

    The numbers which describe the intensity and frequency of the buzz, the formants, and the residue signal, can be stored or transmitted somewhere else. LPC synthesizes the speech signal by reversing the process: use the buzz parameters and the residue to create a source signal, use the formants to create a filter (which represents the tube), and run the source through the filter, resulting in speech.

    Because speech signals vary with time, this process is done on short chunks of the speech signal, which are called frames; generally, 30 to 50 frames per second give an intelligible speech with good compression.

  • Arduino:
  • More thoughts on cheap "reading glasses" version of a hearing aid, based on the average profile?

July 15, 2022: 9:00

General

  • Steve is in Japan; back around 7/21

Biplov

  • Reports on
    1. microphones, and noise reduction
    2. fft library for arduino
    3. Transform x=0-1023 into N octaves, from 250:

      Note that since your stepper is based on powers of 2, x runs over a power of 2 () steps. Maybe a better way to represent the function is this:

      What we've done is create a composition of two functions: your linear function has become an input to an exponential function (base 2), to create equal treatment of octaves (rather than equal treatment of linear stretches of frequencies).

    4. multiplexer for switching between ears

Andy

  • KYNETIC
    • KCV - Kentucky Commercialization Ventures

      Kentucky Commercialization Ventures is Kentucky's unprecedented engine for statewide innovation, created to fuel economic growth and ignite life-changing impact for individuals, businesses, and communities. As a newly created science and technology public-private partnership, we provide resources to the state's public universities and colleges to transform ideas into services, products, processes, startups, and investments supported by intellectual property.

    • KSTC -- The Kentucky Science and Technology Corporation (KSTC) has partnered with the Kentucky Cabinet for Economic Development, the Kentucky Council on Postsecondary Education, the University of Kentucky, and the University of Louisville to lead the development of this statewide collaboration. The KCV initiative is a proud recipient of the 2021 Lab to Market Inclusive Innovation Ecosystems award from the U.S. Small Business Administration.
    • KYNETIC -- Kentucky Network for Innovation & Commercialization (Supporting health-related technology development at Kentucky's public universities and colleges)

      The KYNETIC hub is led by the University of Kentucky (UK), University of Louisville (UofL), the Kentucky Cabinet for Economic Development, and Kentucky Commercialization Ventures (KCV). The long-term goals of the program are to foster commercialization success, economic development and entrepreneurial culture change throughout Kentucky.

      KYNETIC is a competitive grant program, focused on product development to address unmet needs to benefit human health. Any member of the participating institutions in Kentucky is allowed to apply by submitting a short pre-application. Pre-applications will be reviewed as described below and the most promising will be invited to submit a full application.

    • Megan and I had a really good conversation. She assured me that I don't need to worry about someone swooping in and appropriating this for their own nefarious purposes.

    • Here are some links for next steps (from Megan):
      1. LaunchBlue - this is one of the bootcamps you are eligible to attend (at no cost to you)
      2. LaunchIt: we are checking on virtual options for this one; its shorter and slightly different material than the UAccel program (again, no cost to you)
      3. NKU Disclosure Form; as part of the email, you can let them know youve spoken with KCV already if you want.

    • Biplov:
      • Already working with KCV and KYNETIC
      • Time won't allow work in the next year
      • Maybe bring a second student from engineering into the project
      • Work on this through the fall, trained by Biplov; knows a good freshman, and can check in with him.
      • Mechatronics Engineering Technology-- in charge of capstone; Dr. Mahdi Yazdanpour
  • Mathematica
    • It occurred to me that Mathematica may be able to record sounds; it can! So we can create some audio snippets, with high frequencies, and then replay them with the shift.... Here's an example
    • AudioPitchShift: This is a great built-in which does this across the spectrum of a sound. So this can be done in theory, relatively quickly.
    • Loudness: Mathematica seems to use RMS Amplitude,
  • Arduino:
  • More thoughts on cheap "reading glasses" version of a hearing aid, based on the average profile?

July 12, 2022: 8:00

General

  • Steve is in Japan; back around 7/21

Biplov

  • Reports on
    1. JMM poster session dates and details?
      • happens Jan 4(Wed) to Jan 7(Saturday) 2023 at John B. Hynes Veterans Memorial Convention Center, Boston Marriott Hotel, and Boston Sheraton Hotel, Boston, MA
      • deadline for all abstract submissions: sep 13, 2022
      • abstract submission is online; further notice will be available soon
      • two types of presentation: special session and contributed papers - 20 mins
      • speakers and internet in session rooms are not available
    2. Finish tester? => one speaker tester is done
    3. Look into bone transducer and conduction?
    4. Transform 0-1023 into octaves....
    5. multiplexer for switching between ears. => still figuring out multiplexer
    6. Raspberry Pi?
  • Did you get a chance to look at the Music website, or the python stuff?
  • Euler's identity is based on Euler's formula,

    and is the coolest identity in the world:

Andy

  • Mathematica
    • Playing my tester results as notes -- not quite so much bird song at high frequencies!
    • Playing my tester results as notes, and cut-off -- bringing high frequencies down where we can hear.
    • AudioPitchShift: This is a great built-in which does this across the spectrum of a sound. So this can be done in theory, relatively quickly.
    • It occurred to me that Mathematica may be able to record sounds; it can! So we can create some audio snippets, with high frequencies, and then replay them with the shift....
    • Loudness (got these from the Help for AudioIntervals)
      • Stevens's power law: "The general form of the law is
        where I is the intensity or strength of the stimulus in physical units (energy, weight, pressure, mixture proportions, etc.), ψ(I) is the magnitude of the sensation evoked by the stimulus, a is an exponent that depends on the type of stimulation or sensory modality, and k is a proportionality constant that depends on the units used."

        For sound, Stevens found , for "Sound pressure of 3000 Hz tone"

      • loudness according to EBU momentary standard
  • Python:
    • I've got Jupyter working, but not yet pyaudio.
    • Lots of struggles with python
  • Joe is big into the Electret microphone. Says we need to use that. He is also very interested in the colored lights aspect: I think that this might be the YouTube of the Arduino Christmas light show that Joe was referring to (although he talked about "bushes". It comes with code on github for doing the control....

  • More thoughts on cheap "reading glasses" version of a hearing aid, based on the average profile?

July 5, 2022: 9:00

General

  • Steve is headed for Japan; back around 7/21

Biplov

  • Regarding what's new on the hardware front:

Andy

  • Mathematica
    • Playing my tester results to get something somewhat realistic -- but we're still hearing some bird song at high frequencies

      • Again, note the effect of sampling rate.
      • I think that the "birdsong" at the end might be because we're changing frequencies so rapidly that the choice of which frequency to play gets aliased.
    • Smoothers versus spline: how does hearing typically degrade over time?
    • It might be easier if we stick with octave frequencies, or "half octaves" (e.g. multiplied by ); this will make the test results a little easier to interpret. We can play the results using (e.g.) , rather than spline-fitting the frequency. We'd just need to fit the amplitudes. But, from an adaptive level, it gives us less freedom to play.

      In fact, I guess we could just use Log[2,freq/250] as time? Doh!:)

  • Some reflections:
    • I found a very good introduction to the physics of sound and music. In particular, it has a nice discussion of noise:
      • Formally, white noise is a sound with a flat power spectrum; that is, it transmits power equally at all frequencies. It is a mathematical ideal with a representation something like this…
      • Formally, pink noise is a sound with a power spectral density that is inversely proportional to frequency over the range of audible frequencies. Because of this it is also known as 1/f noise (one-over-f noise). We can write this as a proportionality statement.

        We tend to group sounds perceptually into bands that increase by successive powers of 2. A mathematician would call this a logarithmic increase. Musicians call the bands octaves. Audio technicians developed pink noise as a response to this reality.

        A pink noise signal transmits energy equally over all octaves (logarithmic intervals). Let's test this essential characteristic. Compute the power radiated in a frequency band spanning the octave f = (a, 2a).

    • I want to play with this: Audio Processing in Python Part I: Sampling, Nyquist, and the Fast Fourier Transform

    • The big radical idea:
      What if we add a color -- use light! -- to augment the higher frequency? I found this interesting attempt: The Color of Sound
  • On my todo list:

July 2, 2022: 8:00

General

  • Steve's last meeting for awhile (gone from 7/5-7/21)

Biplov

  • Decibel sensor?
      • Arduino and microphone can be used to make a decibelmeter.
  • What's new on the hardware front?
      • 1. potentiometer to control volume => works well
      • 2. potentiometer to control frequency => huge fluctuations, cannot a certain frequency.

Andy

  • Mathematica
    • Steve's new, cool tester
    • My results (uncorrected) (with the use of splines)

      Notice that I've altered the tester a bit more, especially per sampling rate.

    • Note that I scaled back the first volume on the second pass; 1 is a little harsh on some of those frequencies! It might be best to set it to twice (say) the result from the first pass.
    • "sampleRate" -> 64007

      Last time I said I'm stone deaf at 8000. But the sampling rate was 16000, and we're all stone deaf at 8000 with that sampling rate! My idea now is to use a prime sampling rate, but it's interesting to look at the differences if you play back my results at different rates. You can at least SEE dramatic differences, even if you can't necessarily hear them.

    • Smoothers versus spline: how does hearing typically degrade over time?
    • It might be easier if we stick with octave frequencies, or "half octaves" (e.g. multiplied by ); this will make the test results a little easier to interpret. We can play the results using (e.g.) , rather than spline-fitting the frequency. We'd just need to fit the amplitudes. But, from an adaptive level, it gives us less freedom to play.

      In fact, I guess we could just use Log[2,freq/250] as time? Doh!:)

  • Some reflections:
    • I forgot phase last time, in the mathematical description! Yikes! I think that I've got it fixed now: check out the the working document, "Hearing Correction" section.

      For a quick and dirty on phase, consider this page from a course entitled "The Mathematics of Music": Sound Waves

    • The big radical idea:
      What if we add a color -- use light! -- to augment the higher frequency? I found this interesting attempt: The Color of Sound
  • On my todo list:

Steve

  • Adaptive frequency exploration.
  • distortVolume should be saved, rather than "volume".
  • Any chance to do some testing?

June 27, 2022: 9:00

General

  • Steve will be gone from 7/5-7/21

Biplov

  • Purchases -- Yes, the orders came in.
  • Any chance to do some testing?

Andy

  • wiki issues
    • Can't save some file types; I put them elsewhere, and then make links in the wiki
    • Collisions can sometimes be "fatal". Copy wiki code to an editor, if you're nervous, so that you can easily put it back later.
    • There is a history for each page. I found, upon updating the "MediaWiki" software, that I'd allowed anonymous folks to create accounts; so I spent an hour or so yesterday reconfiguring, blocking folks (or bots), and destroying some of their pages. Those users could have tampered with our pages, and caused trouble. Yikes!
  • the typical strategy of hearing aids
  • Mathematica
    • Generated updated results, including corrected hearing results: have a look at my results (with correction)
    • I have also created a spline, to interpolate. This is crude, but turns data into a function.
  • Some reflections:
    • We need to standardize vocabulary:
      1. "volume difference profile" -- "profile"
      2. "frequency shift rule" -- "rule"
      Let's have a look at the working document, "Hearing Correction" section, to see how I'm using terms (and to talk about some math). In particular,
      1. how do we turn the threshold profile function into a weight function for the amplitude of a sound?
      2. how do we think about some of these integrals? (e.g. time varying volume?) I'm sure that some of this is out there, just haven't found it (or looked hard enough) yet....
    • In terms of adaptive testing, my results went (on a scale of 0 to 1) from about .24 to 1 (deaf!). So, to retest, and make finer distinction, we might choose a "distortVolume" function that starts at .24....
    • Should we consider splitting frequencies -- keeping some energy in the original, but passing some to another frequency? Would this sound like an echo chamber, and drive people crazy?
    • We should try to preserve "apparent" energy of sounds.
    • The big radical idea:
      What if we add a color -- use light! -- to augment the higher frequency? I found this interesting attempt: The Sound of Color....
  • Still on my todo list:

Steve

  • Thoughts of adaptive frequency exploration.
  • Any chance to do some testing?



June 24, 2022: 12:30

General

  • Steve will be gone from 7/5-7/21

Biplov

  • Purchases -- The potentiometer and speakers will be delivered today - Ms. Landwehr replied this morning.
  • How will pyfirmata and WaveHC going to work together ???
    • -> As much as I have understood, we cannot run both together because WaveHC is in C whereas pyfirmata needs to have a specific script uploaded to arduino
    • -> we can use switch for modes to solve this problem: whiteboard
  • Thinking about the "circular buffer" -> there are open source git repos for creating it, both in python and C/Arduino
    • We can't write on the SD card in wave shield, so circular buffer can not be used
  • Storing massively many wave files on an SD card -> this is still an option: doing the discrete adaptive frequency play
  • pyAudioAnalysis
    • It will let us recognize background noise and human voice by the use of CNN.
    • How to obtain just the human voice??

Andy

  • First thing:
    lots of things in life improve iteratively, by bootstrapping. Steve, this Mathematica tester has been in inspirational -- I have seen things a little more clearly. Thanks! I think that you all might try it, too, and see what your results look like. One advantage for us if you do it Steve is we have a fresh audiology report on you, and we can try to regress the two test results to figure out the decibel translation....
  • wiki issues
    • Can't save some file types; I put them elsewhere, and then make links in the wiki
    • Collisions can sometimes be "fatal". Copy wiki code to an editor, if you're nervous, so that you can easily put it back later.
    • There is a history for each page. I found, upon updating the "MediaWiki" software, that I'd allowed anonymous folks to create accounts; so I spent an hour or so yesterday reconfiguring, blocking folks (or bots), and destroying some of their pages. Those users could have tampered with our pages, and caused trouble. Yikes!
  • Mathematica
  • Some reflections:
    • We need to standardize vocabulary:
      1. "volume difference profile" -- "profile"
      2. "frequency shift rule" -- "rule"
    • In terms of adaptive testing, my results went (on a scale of 0 to 1) from about .24 to 1 (deaf!). So, to retest, and make finer distinction, we might choose a "distortVolume" function that starts at .24....
    • Should we consider splitting frequencies -- keeping some energy in the original, but passing some to another frequency? Would this sound like an echo chamber, and drive people crazy?
    • We should try to preserve "apparent" energy of sounds.
    • Here's today's big radical idea:
      If we make our rule function recursive, e.g. if I am deaf to frequency , , then we are aliasing, essentially, and losing information. What if we add a color -- use light! -- to augment the higher frequency?
  • R libraries for sound analysis: I’d said I’d look into R for testing -- haven't yet got to that.

Steve

  • Thoughts of adaptive frequency exploration.
  • How do left/right ear disconnects impact the listener?

June 21, 2022: 9 am

Biplov

  • Purchases -- did they come in? => Our Arduino boards have been delivered, but the potentiometers and speakers still have not.
  • Waveshield Voice changer: not required for testing phase, but might be useful in Aid
  • Waveshield:
  • Biplov mentioned the idea of restricting frequencies/harmonics, essentially as a noise-reduction strategy.
  • In terms of the tester, Biplov's current idea is to create a slew of small wav files that each contain a pitch and a volume, and then allow the user to adjust potentiometers to set frequency and then allow the volume knob to increase until the user can hear the pitch; record that, and continue.

Andy

Steve

  • Steve will be gone from 7/5-7/21
  • Thoughts of adaptive frequency exploration.
  • How do left/right ear disconnects impact the listener?

June 17, 2022: 9:45 am

Biplov

  • Mathematica => played and understood with your provided code
    • Volume control issues? => wave shield
  • Purchases -- did they come in? => I haven't received any email back from Mrs. Landwehr. Emailed to get the status
  • Background report? => Apologize, I'll have all of it on by tomorrow


  • Wave Board (standalone): https://www.adafruit.com/product/2220
    • 3-5.5 v DC
    • 2 MB or 16 MB flask memory
    • stereo support (44.1KHz 16-bit stereo)
    • 11 triggers
    • also comes with a version of stereo speaker amplifier (16 MB)



  • Wave Shield: https://www.adafruit.com/product/94#technical-details
    • Only compatible with ATMega328 based Arduino
    • It Can play up to 22KHz 12-bit uncompressed audio files of any length and momo wave files of any size
    • Need to buy SD card separately
    • Our ordered Arduino board is based on ATMega328, so it is compatible with wave shield


Andy

  • Mathematica
  • Python and R libraries for sound analysis: I’d said I’d look into analysis for python and R: here are some early looks

    Steve

    June 13, 2022: 9 am

    Biplov

    • Mathematica?
      • Under control
    • Norsemathology?
    • Purchases....
    1. 2 - arduino Uno starter kit
    2. 4 - potentiometers
    3. 10pcs pack - minature full range speakers
    • Fill Background in report

    Andy

    • Mathematica
      • Play
      • Sample rate
      • Audiograms -- how they work. Why do they choose those frequencies?
    • gizmos suggested by Joe Longstaff:
      • Wave board?
      • Electet

    Steve

    • Unavailable
    • We need an image of his (relatively normal) audiogram

    Meeting, Tuesday, June 7th, 2022: 9 am

    Biplov

    • Biplov was not feeling well, and couldn't attend.

    Andy

    • Mathematica
      • Play
      • Sample rate

    Steve

    • Hearing test!

    Meeting, June 2nd, 2022

    1. We need a better name. Retraining your brain? I've switched to HearWell....
    2. Andy's items
      1. Bring Steve up to speed
      2. A couple of my audiograms....
      3. starting the report
      4. some imaginings, to share with Steve — thinking like a bat, for example. Biplov talked about stereo mics, and Andy suggested three mics might work (three satellites are necessary to geolocate a position on the Earth).
      5. Sound file examples we might want to use. (Didn’t get to this)
      6. Andy asked Biplov to look into two things:
        1. Publication options, and
        2. Intellectual property issues.
    3. Biplov items:
      1. regarding prices of apps and such. Andy has an email from Brooke offering up to $500 more, if necessary.
      2. Mathematica activation? Still not there….
      3. wiki access? Ditto: not yet.
      4. Biplov had several questions, some of which were related to
        1. technical definitions of hearing (e.g. Tone versus Timber, pitch versus frequency, color; 12-tone scale)
        2. nerve and other hearing damages — what are the mechanisms for causing one to lose hearing? Questions for Dr. Sharma…..
          1. interference? Collisions?
    4. Steve:
      1. Mathematica commands
      2. We talked about how Peter’s work might be helpful. Also whether it would be better to switch to python, say, or R — something public domain. Biplov described how the Arduino connects with Mathematica, but mentioned that he would look into what options exist for python or R.


    Meeting, May 30th, 2022

    1. Andy's items
      1. Dr. Sharma's notions — in working document
      2. Some references — from homepage
      3. Mathematica tools — in todos
        1. In particular, it seems that the periodogram function does the Fourier Series Analysis that we’re going to want to do.
      4. IRB approval — I need to get on this! Add this to todos….
      5. A role for Thad?
    2. Biplov's report
      1. Biplov reported on the apps and the device that Dr. Sharma had suggested that we look into. There are a couple of possible winners.
      2. Biplov needs a Mathematica activation code
      3. We got a norsemathology password sent to Biplov
      4. Biplov’s capstone allows him to get some training from Engineering Physics, if necessary.
      5. Biplov will be available to purchase items in the department, to make it easier on Joni or Kimberly.