Webinar: Engineering Smart Glasses Audio for the Real World.

Webinar agenda

• The smart glasses audio challenge
  Explore how system engineers and AI/ML engineers can use the same virtual product and acoustic environments to support development, testing and training.
• Create and characterise the virtual device
  Import a glasses design, configure device and wearer variants, and define the microphone array. Generate device-related transfer functions, understand how they differ from head-related transfer functions, and explore wearer voice capture.
• Build and scale acoustic scenes
  Place the characterised device in representative environments, add speech and noise, and render multichannel recordings. Scale one scene across wearers, rooms, source positions, noise conditions and signal-to-noise ratios. A separate receiver-only demonstration will illustrate movement.
• Generate and inspect ML-ready data
  Turn the scene population into structured datasets with device-specific recordings, clean references and acoustic metadata. Inspect and filter the data to isolate challenging conditions and target gaps in training coverage.
• Evaluate system performance
  Pass the generated recordings through an external audio-processing algorithm and assess the output using metrics such as word error rate and voice quality. Connect results to acoustic metadata to identify which conditions cause performance problems.
• Explore next steps with Treble
  Review the complete workflow, preview relevant roadmap features and explore ways to get started through a web-based demo, trial, no-install SDK environment or guided onboarding.