CI Hackathon 2020

Crowdsourced science to help cochlear implant wearers hear better

FIRST COCHLEAR IMPLANT HACKATHON DECLARES WINNING TEAMS; ALGORITHMS PERFORM BETTER THAN GOLD STANDARD

The Cochlear Implant Hackathon is a joint effort between Advanced Bionics, University of California San Francisco, and University of Minnesota with a goal to inspire members of the general public to improve cochlear implant (CI) sound processing. Advanced Bionics, one of three FDA-approved cochlear implant manufacturers, sponsored the contest and provided a “gold standard” reference implementation of a fully-featured cochlear implant sound coding strategy. Advanced Bionics and the University of Minnesota further provided a software framework for contestants to develop and test their own algorithms, including an acoustic simulation of the sound percept produced by a CI in order for normal-hearing listeners to be able to optimize and ultimately judge the sound quality of the CI sound processing algorithms.

After the development phase of the competition, each algorithm was crowd-judged for its performance in four different sound categories: natural speech, speech in a noisy environment, simple words, and music. A virtual competition format encouraged international participation with contestants from across five continents. Top tier entries represented academic institutions, commercial entities, and individuals. Winning teams for the Cochlear Implant Hackathon were announced this week, with awards of $5000 for first prize, $2000 for second, and $1000 for third. Bonus awards of $250 were awarded to other top four teams in each category for performing better than the “gold standard”. Top three teams will have opportunities to collaborate with Advanced Bionics to integrate novel strategies with current sound processing techniques.

Winning Teams

The top three teams which performed well in each of the four categories are:

  1. Pepino-Barchi - Leonardo Pepino and Germán Barchi, independent researchers with a background in acoustic engineering, based in Buenos Aires, Argentina
  2. Googlears - Richard F. Lyon, Google Research; Mihajlo Velimirović, Google Research; Samuel J. Yang, Google Research; Scott Wisdom, Google Research; Pascal Getreuer, Google Research; Chet Gnegy, Google Research; Sagar Savla, Google Research
  3. Binding Hack - Ravinderjit Singh, M.D. / PhD Candidate, IUSM/Purdue University; Andres Llico, M.S., PhD Candidate, Purdue University

Honorable mentions for exceeding gold standard performance in at least one judging category go to the following teams (in alphabetical order):

16-SAMurai: Tamás Harczos, (1) Fraunhofer Institute for Digital Media Technology IDMT, Ilmenau, Germany, (2) Auditory Neuroscience and Optogenetics Laboratory, German Primate Center, Göttingen, Germany, (3) audifon GmbH & Co. KG, Kölleda, Germany; András Kátai, Fraunhofer Institute for Digital Media Technology IDMT, Ilmenau, Germany

Cochlear Ethereal Audio: Alan Kan, Ph.D., Macquarie University; Qinglin Meng, Ph.D., Lecturer, South China University of Technology; Huali Zhou,Graduate student, South China University of Technology; Zhenyu Guo, Doctoral student, South China University of Technology; Yu Lan, Graduate student, South China University of Technology; Junming Li, Undergraduate student, South China University of Technology; Nengheng Zheng, Associate Professor, Shenzhen University; Yuyong Kang, Graduate student, Shenzhen University; Xi Chen, Graduate student, Shenzhen University; Yupeng Shi, Researcher, Tencent Ethereal Audio Lab; Wei Xiao, Principal Researcher, Tencent Ethereal Audio Lab; Shidong Shang, Head of the Tencent Ethereal Audio Lab

Haphazard Hamming: Max Jiam, Undergraduate Student at Carnegie Mellon University

Southeastern-3: Sanichiro Yoshida, Ph.D., Professor of Physics, Southeastern Louisiana University; Anthony Calmes, Graduate student, Integrated Science and Technology, Southeastern Louisiana University; Conor McGibboney, Graduate student, Integrated Science and Technology, Southeastern Louisiana University

Steel City Sounds: Keiko Gutierrez, B Mus., McMaster University; Daniel Shields, B Eng., McMaster University; Brendan Tao, McMaster University; Larissa Taylor, MASc, McMaster University; Michael Wirtzfeld, PhD, McMaster University; Melih Yahli, MASc, McMaster University

What are cochlear implants?

Cochlear implants are electrical devices that restore hearing to people who are born without hearing or lose their hearing over time.


Why hack the cochlear implant?

Cochlear implants are devices that allow some deaf people to hear sounds by overriding the innate hearing mechanisms of the ear. While cochlear implants can restore or improve hearing, users may experience difficulty, especially with hearing speech in a noisy environment and enjoying music.

The CI Hackathon was born from an idea that getting fresh perspectives could lead to innovation and improvement for current solutions. The CI Hackathon ranked entries based on their performance in four sound categories: a series of three unrelated one-syllable words, natural speech, speech in a noisy background, and music across a wide range of styles.

A completely virtual environment

Traditionally, hackathons are held as in-person, weekend-long contests in which participants regularly pull all-nighters. What made this hackathon unique was that it was conducted entirely online via a web-based application, as necessitated by the Covid-19 pandemic. Asynchronous delivery via an online application enabled geographically diverse teams to participate with fewer time constraints. This constraint ultimately proved to be beneficial since improving the quality of cochlear implant algorithms is a difficult task, especially for contestants who had never worked in the field. Contestants were given eight weeks to learn how current cochlear implants and CI algorithms work, implement their own changes to the processing pipeline, test algorithms with training data, and submit entries. From December 2020 through January 2021, teams refined their research, after which final submissions were judged in three rounds, with only top contestants progressing to each successive round. The first two rounds were judged using crowd-sourcing by the participants and the general public listening to entrants’ sounds online and rating them. The final round was an interview with the CI Hackathon judging panel.

Concluding Remarks

The CI Hackathon achieved its goal of sparking an interest in cochlear implant research and bettering the hearing of those who use cochlear implants. The organizing team intends to share results of this hackathon and make them publicly available through publication in conjunction with the participants that produced top results.

This year was the inaugural year for the CI Hackathon, and because it was such a successful endeavor, the organizing team hopes to make this a recurring event.