Wisear’s mission is to build the future of human-machine interfaces
It’s an ambitious goal that requires a lot of R&D and scientific or technical investments, and we could easily lose sight of practicality and usefulness while pursuing it.
To stay grounded and develop a technology that will really enhance people’s lives, one of our core values is to be science-first, which means always making sure that we keep up with the most recent expert research to develop the world’s best performing neural interface.
We don't want to limit the use of our tech to well-controlled lab settings. Our vision requires us to adopt the philosophy and working methods of a full-stack deep tech company. We aim to create a neural interface that provides high-speed, private and accessible controls that can be used by anyone, anywhere, at any time, regardless of the environment—whether operating heavy machinery or on a bumpy bike ride up in the mountains. We turn our lab research into reliable products that can withstand the demands of real-life situations!
Our activities have always been guided by two principles: science-first and real-life proofness. We take pride in the fact that our efforts and expertise are now recognized by our peers and leading R&D teams worldwide.
Earlier this year, we were asked to present Wisear and our technology to the Microsoft Research team, an invitation we gladly accepted from Ivan Tashev. This was a unique opportunity for us to get direct feedback. We believe that the future of XR relies on the cooperation of all players in the ecosystem, not just isolated initiatives, even if they are breakthrough innovations (check out our latest article on this).
Our CEO, Yacine ACHIAKH, recently took a quick trip to Seattle, just two weeks after the exciting CES event that’s always buzzing with energy. He was thrilled to discover the inspiring Microsoft Research Lab offices, where developers can directly code their own arcade games on a vintage machine at the entrance of the building!
While there, Yacine and Alain Sirois met with the BCI engineers, who represent over 1000Bn neurons activated 🧠, and shared an overview of our R&D achievements over the past three years. A golden opportunity to discuss two of our main efforts in hardware and AI with the Microsoft Research team!
Hardware - our sensor design decisions
1. Electrodes play a crucial role in our technology: they collect bioelectrical signals and convert the ionic currents flowing through the body into electric currents that can be processed by our electronics. This is a key element for our tech, because the better the electrodes, the better the signal quality, which then allows us to maximize the performance of our gesture detection.
2. To design our electrodes, we first identified several constraints that needed to be played around with:
Electrode-skin impedance (which prevents electric currents from flowing from the skin to the electrode) and the impedance mismatch between electrodes need to be minimized.This is crucial in maximizing the signal-to-noise ratio and overall signal quality.
It's essential to keep the cost of the electrodes down to ensure that it has a negligible impact on the BOM (bill of materials, or overall cost of components) and doesn't become a blocker for integration with our partners.
The electrodes need to be biocompatible, comfortable and easy to clean, to comply with all regulations and provide a positive user experience.
3. Once we identified the constraints, we ran experiments and considered different factors until we found the best match:
We tested a number of materials and approaches (metals, composite polymers, conductive coatings, intrinsically conductive polymers, capacitive electrodes), eventually sourcing a conductive polymer material that was off-the-shelf, affordable, biocompatible, soft, and easy to clean.
We tested different shapes, sizes and positions for the electrodes to maximize the contact area and ensure stability across various usages and environments. Ultimately, we decided to craft electrodes that could replace the soft parts of typical earbuds, providing superior comfort while also lowering the overall BOM by replacing parts of traditional earbuds.
4. This puts us in a really good position, as we can now turn the next generation of earbuds into neural sensing devices while barely impacting the BOM or the look and feel that has been refined across multiple product generations.
AI - How we built our data and signal processing capabilities
1. To create human-computer interfaces that leverage neural signals, access to large amounts of high-quality data to train AI models is key.
2. This comes with its share of challenges:
As a relatively new field, publicly available data is still very limited. It's particularly challenging to build a solution that doesn't require calibration and works for everyone in any environment, as we've done, without access to a diverse pool of data that includes subjects and environments.
Since the associated hardware is still in its early days and rapidly evolving, the recorded signals are often noisy, riddled with artifacts and vary across different generations of hardware, hence the need for sophisticated signal processing algorithms.
Finally, to be a viable option for the next generation of human-machine interfaces, Wisear's tech has to run extremely fast to provide high-speed controls, while consuming minimal power to avoid draining device batteries and overburdening computational loads.
3. We kept these challenges in mind from the beginning, which allowed us to build our roadmap and strategy accordingly:
We put a lot of effort into collecting diverse proprietary data early on, which resulted in significant datasets that we can leverage to develop our algorithms. We also developed proprietary data augmentation techniques to further increase the quantity and diversity of our data pool.
At the same time, we applied advanced signal processing and machine learning techniques to eliminate noise and artifacts from recordings, and to standardize data coming from different recording setups.
Lastly, we've been developing our own optimization techniques in house to ensure that our gesture detection algorithms have extremely low latency (and will continue decreasing) and can be embedded on chips with limited memory and computational power.
4. These elements actually give us a strong competitive edge:
Data collection is a lengthy process, but we've made significant strides in acquiring a substantial pool of proprietary data that’s fine-tuned to our needs.
Most of the BCI/HCI applications that we see today still run on laptops or smartphones, are slow and require individual calibration. This can be tedious and may not work for everyone. In contrast, Wisear's technology is designed to run on earbud chips, making it instantly accessible to all users without the need for individual calibration.
Collaboration within the ecosystem will drive the BCI revolution
This session was rich in valuable insights that reinforce the direction in which Wisear is headed. Here's a summary of the key learnings from these discussions with the Microsoft Research Lab BCI teams:
It’s been three years since our creation, and Wisear and our R&D efforts are gaining recognition in the neurotech industry, giving us the confidence to continue pushing forward with our vision and driving progress within our ecosystem.
Our hardware and software approaches seem to be aligned with the roadmap and convictions of the Microsoft Research Lab teams. We can, therefore, double down on our R&D efforts with a clear path forward.
Collaboration is a key factor that we can leverage to advance the BCI ecosystem. Similar to the recent acceleration and democratization of AI made possible by partnerships with 1,000+ labs, disruptive neurotech changes will come from strong interdisciplinary and global collaboration within the industry.
We're proud of our collaboration with Microsoft Research and thrilled to be embarking on the next steps of this BCI adventure with them.