Thanks Dr. Launer for that answer!
Regarding training data for AI: are there considerations within the HA industry to collaborate and create a shared dataset that all participating manufacturers can use? Or maybe even an open source dataset?
Thanks Dr. Launer for that answer!
Regarding training data for AI: are there considerations within the HA industry to collaborate and create a shared dataset that all participating manufacturers can use? Or maybe even an open source dataset?
Process size: at Sonova we use advanced micro-electronics process sizes which are available for the chips that need to be implemented, so various process types for different chips in the hearing aid, DSP, AI-chip, wireless radio, memory… Moving forward, process size certainly helps to further reduce size and power consumption but the technology becomes more challenging to master and also more costly to develop.
Dr. Launer
Can you reveal who your manufacturing partner is? A while ago Oticon and TSMC put out press releases about their cooperation and obviously TSMC is at the forefront of process technology.
I think the really critical thing for codergeek is not necessarily the form factor, but that the Infinio CROS RIC isn’t compatible with the BTE. This has left a lot of users with severe loss in one ear and no aidable hearing in the lurch. Other manufacturers have CROS solutions for UP BTEs, but they don’t have good connectivity options for Roger with is also pretty critical for this population.
Does AI in hearing aids mean that the pace at which Phonak can improve performance (power consumption, speech in noise performance, etc) actually increases? Meaning that hearing aids will improve more in the next 5 years than they did in the last 5 years?
Are there any plans of bringing in a Deepsonic AI Chip to a Roger device, so other Phonak users can gain from the Speech in Loud noise improvements?
Is your AI trained on few specific languages or independent of any languages!
Meaning, is it only trained on English, Spanish, Scandinavian languages!!! or it has nothing to do with spoken language??
Thanks
Thank you @Neville. You are 100% correct. They still haven’t answered that part of the question. When the Phonak Naida UP BTE FORMAT is coming out. There currently is no Naida UP BTE FORMAT AVAILABLE FOR PURCHASE and I want to know exactly when this is happening since my Naida UP BTE B90 plus CROS died and I am stuck in limbo with Oticon until I can purchase new Phonaks which are not available.
Another AI training data question: how do you obtain this data? Is it people going out into the real world with microphones or is it created in a lab environment?
DEEPSONIC processing power: Deepsonic is a specific Neural Processing Unit running an AI model with 4.5 Mio parameters with more than 7 GOPS, remeber, that has to run off a hearing aid battery.
Dr. Launer
Are specifics considered proprietary? Thanks! (This is in reference to process size)
Great question, not currently but possibly in the future.
Steve
Phonak did reveal that the AI chip runs at 50MHz… Which is surprisingly low, but it’s in line with for example OnSemi’s Ezairo chip which Starkey uses.
Those high clock rates that we are normally used to are simply not possible with the extremely limited power budget that HAs have.
It might not actually be in the works for this platform coder. In which case you’ll have an answer but you will be disappointed, but you still have to be nice to them or they won’t come back.
It is trained to many languages, not yet to specific languages
Thanks for this question. Having AI in a hearing aid is like having an engine in a car. It strongly depends on the performance of that engine. The size, architecture, and training matters alot! Other brands use the DNN to classify the sound scene or for light noise reduction. Phonak uses the DNN to extract, enhance, and integrate the speech. This powerful feature is known as Spheric Speech Clarity and instantly separates speech from noise, delivering an unprecedented 10dB SNR.
Can Phonak’s AI system can help with understanding higher pitch speech in noise for someone with good hearing up to 2kHz but with very poor higher frequency hearing partly due to tinnitus?
Can you be more specific? Which languages? And what’s the breakdown among those languages?
Do you have anything in the works using AI which can help severe to profound hearing loss, i.e. anything before getting a cochlear ear transplant?
@Neville - I’m being nice. I am simply using caps to emphasize that I would like a direct answer to my question which so far has not been provided. All I want to know is when the Phonak Naida BTE Ultra Power plus Cros Ultra Power form factor is being released. Soon? September? December? January? It does no good for me to tell me what’s already available because it doesn’t answer my question.