eWEAR Seminar: AI insights from wearable sensors_Stanford University

Tonight, I had the opportunity to remotely attend a seminar held at Stanford University, where two insightful presentations shed light on advancements in wearable sensors.

Wearable Sensing and Generative Deep Learning for Gait Dynamics Assessment

Speaker: Tian Tan, Postdoctoral Researcher in Radiology, Stanford University

Key Applications of Gait Dynamics Assessment:

• Injury Prevention: Reducing the risk of physical injuries through better understanding of movement patterns.
• Disease Assessment: Early detection and monitoring of conditions like Parkinson’s or diabetic neuropathy.
• Mobility Assistance: Enhancing assistive devices for improved functionality.

Traditional Methods:

• Marker-Based Motion Tracking and Force Plates
  These systems are accurate but come with significant drawbacks:
  • Confined to controlled environments.
  • High operational costs.

The Shift to Wearable IMUs:

• Inertial Measurement Units (IMUs) offer portability and affordability, enabling data collection in real-world environments such as homes and outdoor settings.

Challenges in Gait Analysis Models:

1. Data Bottlenecks:
   • High-cost experiments often result in small, limited datasets.
2. Model Architecture:
   • Dependence on strict input-output pairs limits adaptability.

Key Studies Shared:

1. Self-Supervised Learning:

   • Enables models to learn without labeled data, drastically reducing dependency on costly, annotated datasets.
   • Enhances adaptability and scalability.

2. Generative Diffusion Models:

   • Compared traditional end-to-end models with generative diffusion approaches.
   • Used diffusion models to estimate external forces and predict gait modification responses—eliminating the need for expensive experiments.

---

MEMS & Sensors in the Metaverse

Speaker: Sneha Kadetotad, Engineering Manager, Motion Sensors, Meta Reality Labs

From Text to Immersion:

Meta Reality Labs is spearheading hardware and software innovations for the Metaverse, focusing on:

• Virtual Reality (VR)
• Wearables

Current Applications of MEMS & Sensors:

1. Hand & Body Tracking:

   • Sensors: Optical depth sensors, machine vision cameras, human vision cameras.

2. Eye & Face Tracking:

   • Sensors: Machine vision cameras.

Challenges in AR/VR Development:

• Capabilities & Performance: Delivering high-fidelity experiences with an advanced feature set.
• Wearability & Social Acceptance: Ensuring devices are comfortable, inconspicuous, and have extended battery life.
• Affordability: Balancing innovation with cost-efficiency.

Emerging Innovations in MEMS & Sensors:

1. Technological Advancements:
   • Development in MEMS, CMOS, specialty technologies, and new materials.
2. Platformization:
   • Advanced packaging and integration.
3. Intelligent MEMS & Sensors:
   • Pioneering smart functionalities for next-gen applications.

New Features & Experiences:

• Health & Wellness Monitoring: Leveraging sensors for continuous health tracking.
• Force Feedback: Improving the tactile realism of interactions in virtual environments.
• Active Displays & Optics: Advancing visual fidelity for AR/VR devices.

Meta is now shifting focus toward building complete platforms, aiming to redefine AR/VR experiences.