/media/bill/HOWELL_BASE/Projects/DrumLib - kid science event/DeLiang Wangs Deep Learning hearing aids.txt www.BillHowell.ca 20Sep2017 initial **************************** PERCEPTION AND COGNITION - DEEP LEARNING HEARING AIDS DeLiang Wang is a professor at Ohio State University in Cincinatti, a recipient of Ohio State's 2014 Distinguished Scholar Award. He is a past President of the International Neural Network Society, of which I am a member, and he is also a co-editor in Chief of the societie's Neural Networks journal, as top publication in the field. He is well-known for his work on auditory scene analysis, and I've known him personally for well over a decade. DeLiang and his colleagues have applied Deep Learning Neural Networks to hearing aids, and their initial results are very promissing, so much so that they have a collaboration with Starky Hearing products, a leading manufacturer of high-quality hearing aids. Hearing aids do not work as well as many would like. For example, my father rarely uses his hearing aid for the same reasons that many others do not - it's hard to hear with them, especially in noisy environments with many conversations going on at the same time. The "Cocktail Party problem" is one example of this. Think of the times that you are with a group of people, all of whom are talking a t the same time, and there is other noise in the room as well. If someone speaks your name, you can magically pick that out of the sounds all around even when you are focussed on another conversation. But as the scene becomes noisier, it becomes more and more difficult to pick out streams of conversation. I have always had this problem in restaurants and bars, as friends of mine carry on conversations that I can't even make out. DeLiang Wang has posted several sound tracks on the web, and he also sent additional material to me for this presentation. As a simple experiment, listen to the initial part each of the short recordings that follow, and write down what you think is being said. Don't worry if this is hard to understand, these are tough examples. Now listen to the sound after it has been processed by DeLiang Wang's Deep Learning Hearing Aids. There is something very important to note about the results. The ears can hear the initial, untreated sounds, but these are not effectively understood for these examples. Your systems of perception are extremely powerful at extracting information from noisy [sound, visual, taste, touch, smell] environments, but problems with your [ears, eyes, taste buds, nose, and fingers] can make it more difficult to understand what is sensed. Processing of that information in the brain also plays a big role, and aging and other conditions can greatly affect cognition, or the understanding of what is sensed. **************************** REFERENCES : https://engineering.osu.edu/news/2017/01/appyling-machine-learning-engineering-professor-aims-reinvent-hearing-aid https://www.youtube.com/watch?v=tBNpglPHQsY https://spectrum.ieee.org/consumer-electronics/audiovideo/deep-learning-reinvents-the-hearing-aid Posted 6 Dec 2016 | 16:00 GMT By DeLiang Wang http://spectrum.ieee.org/ns/audio/hearing1116/Police.128.mp3 http://spectrum.ieee.org/ns/audio/hearing1116/Cold.128.mp3 http://spectrum.ieee.org/ns/audio/hearing1116/pie.128.mp3 http://spectrum.ieee.org/ns/audio/hearing1116/bed.128.mp3 The above examples are in the youtube # enddoc