A love story from Eriksholm

As we enter 2019, Senior Director Uwe Hermann reflects on the developments in AI and sensors he has seen over the years - and declares his love for Eriksholm.


I want to tell you two small secrets, which you will not find in my CV.

Since I was 16, my passion has been dancing. Ball room dancing or ‘Standard Dance’, as it is called in Europe. Among the 10 different dances which are part of the Standard Dance program, I like Rumba the most. It is a love dance, where the role of the male dancer is rather in the background. His main task is to make his partner feel good, to support her performance, and to make her glitter and shine in the spotlight.

Over the past 30 years as a manager, I learned that good management is like dancing a good Rumba. It is about your love and passion for your team, your stakeholders, and customers. Management is best, when it helps the team to perform at its optimum, to sparkle and glitter in the spotlight, and to receive the applause of the customers. I don’t trust managers who want to be stars themselves. The best managers I met over my more than 3 decades in industry were the humble and modest ones.

And yes, I love Eriksholm; the team and the passion for improving life for people with hearing loss.

The second secret is about AI, sensors, and sensor fusion, and it starts with a dark spot in my CV. In my young days as an engineer, back in the eighties, I unintentionally came to work in a military project for a short time. The project taught me three lessons: (1) not to be cheated into something I don’t want by a manager who considers himself a star, (2) sensor fusion is much more powerful than you might believe, and (3) Artificial Intelligence (AI) had huge promise but did not yet work in the eighties.

The project of this time was about fusing a radar and infrared sensor in the warhead of fire-and-forget missiles. A radar signal alone could easily mistake a heap of scrap for a tank and an infrared sensor alone could interpret a manure heap as the heat signature of a tank. However, both sensors together had a frightening detection rate of almost 100%.
The other lesson I learned was that the AI technology of the eighties was insufficient: rule-based programming worked with the computers of this time but failed in principle to fulfill the promise of AI. Neural networks had a huge theoretical promise but did not (yet) work with the computer power of the eighties.

Now, more than thirty years later, we put AI and sensors into hearing aids, and - I just love it – it works miracles. The ear is a pole position for biosensors like EEG, Infrared, motion, and more. Fusing these signals with AI creates completely new possibilities, not only in hearing healthcare but also in general healthcare.

Just one short example on AI: last year I wrote about the fast progress in AI. At the time, we were in the middle of an exciting project where we looked at Deep Neural networks for separating two known voices. Segregating multiple, unknown voices seemed far away, maybe five or more years into the future. Today, we have a working prototype in the lab, which is capable of exactly this challenging task of segregating multiple unknown voices. One big step towards the solution of the so-called ‘cocktail party syndrome’. Again, it is a true love story to see the passion of Eriksholm, our academic partners, and colleagues in William Demant seeking solutions for those who are hard of hearing.

For me personally it was a long and happy journey from this military project of the eighties to today’s healthcare.

“Love is all you need”.

Read news from Eriksholm Research Centre here.