Touchless User Experience

The integration of the Syntiant® NDP creates a "Touchless" environment where the device anticipates user needs through natural interaction. This section focuses on the two primary pillars of this experience: high-accuracy voice triggers and motion-based kinetic controls. By moving these tasks to the NDP, the system achieves an "Instant-On" feel that is impossible with software-based solutions running on a general-purpose CPU.

• Always-On Wake Word Detection: The NDP is optimized for "Near-Field" and "Far-Field" voice triggers. It utilizes a deep neural network to distinguish the specific phonemes of a wake word (e.g., "Hey Syntiant") from hours of background chatter. To ensure the system is both sensitive and reliable, we calculate the Trigger Sensitivity (S) against the background noise floor.

Formula: Trigger Sensitivity ($S$)

$$S = \frac{P(\text{Trigger} \mid \text{Signal})}{P(\text{Trigger} \mid \text{Noise})}$$

(where a higher ratio indicates a model that is more resilient to "False Triggers" in loud environments.)

• Voice ID and Secure Access: Beyond simple word recognition, the NDP120 can perform "Speaker Verification." It extracts unique vocal tract features to ensure that only an authorized user can trigger sensitive commands (e.g., "Unlock the door"). This is validated by measuring the Equal Error Rate (EER) of the biometric match.

Formula: Biometric Match Distance ($D_{\text{bio}}$)

$$D_{\text{bio}} = \sqrt{\sum \left(v_{\text{user}} - v_{\text{stored}}\right)^2}$$

(where v is the embedding vector representing the user's unique voiceprint.)

• Gesture-Based Control for Hands-Free Environments: In scenarios where voice is not ideal—such as a sterile operating room or a loud workshop—the NDP processes data from 6-axis IMUs (Inertial Measurement Units) to recognize gestures like a "Double Tap" in the air or a "Wrist Flick."

• Touchless Time Series Model: The system maps kinetic energy patterns to specific UI actions. To ensure smooth transitions, we calculate the Dynamic Time Warping (DTW) distance to match the real-time gesture against a library of reference movements.

Formula: Gesture Matching ($G_{\text{match}}$)

$$G_{\text{match}} = \min \sum_{k=1}^{K} d(w_k)$$

(where d(wk) represents the cost of aligning the temporal sequences of the user's movement with the gesture template.)

💡 The "Hysteresis" logic in Gestures

"When implementing gestures, always include a small amount of 'Temporal Hysteresis.' Don't trigger the action on the very first frame of a hand wave; wait for the NDP to confirm the movement across 3-4 consecutive windows to avoid accidental triggers from natural body movements."

---