Abstract
We present a theoretical analysis of a perceptual coding approach for networked telepresence and teleaction. Our so-called Weber quantizer is based on Weber’s Law and can be used in haptic data communication to eliminate changes which cannot be perceived by the human operator. The main advantage of this approach is its ability to significantly reduce packet rates without introducing algorithmic delay, ensuring stability in remote robotic control.
Introduction
In networked telepresence and teleaction systems, haptic feedback—the transmission of force and tactile information—is essential for a human operator to interact effectively with a remote environment. However, haptic devices typically require high sampling rates (often 1 kHz) to maintain transparency and stability. Transmitting every sample as an individual network packet creates significant overhead and risks network congestion.
Traditional quantization methods focus on reducing the bit-depth per sample. In contrast, the primary bottleneck in telepresence is often the packet rate. This paper introduces the Weber Quantizer, a perceptually-motivated compression technique that adaptively reduces the temporal resolution of haptic data streams by leveraging human psychophysics.
The Weber Quantizer
The Weber Quantizer is grounded in Weber’s Law, which states that the "Just Noticeable Difference" (JND) of a stimulus is proportional to its initial intensity ($\Delta I = kI$). By applying this principle to haptic signals, the quantizer only transmits a new update when the signal change exceeds the human perception threshold.
Mechanism and Deadbands
The quantizer operates by maintaining a "deadband" around the last transmitted value. If the current signal $q$ remains within the range $[q_{last} - k|q_{last}|, q_{last} + k|q_{last}|]$, no packet is sent. This creates a signal that "steps" only when a perceptible change occurs.
Key advantages of this approach include:
- Adaptive Temporal Resolution: The update rate decreases during periods of steady or imperceptibly changing signals.
- Zero Algorithmic Delay: Decisions are made based on the current sample and the last transmitted value, avoiding the buffering delays common in other compression schemes.
- Perception-Relative Error: The precision of the signal scales automatically with its magnitude, maintaining high fidelity where it matters most to the human operator.
Theoretical Analysis and Results
The authors provide a rigorous mathematical derivation of the Mean Squared Error (MSE) for a signal following a uniform distribution. They successfully derived a closed-form expression: $e = \frac{k^3 a^2}{12}$, where $k$ is the perception threshold and $a$ is the distribution width.
Validation
To verify the analytical model, the authors conducted simulations using synthetic uniformly distributed sequences. The results demonstrated a near-perfect match between the theoretical predictions and the simulated data across various parameters.
Performance and Savings
While theoretical savings for purely random signals were modest (approximately 5% for $k=0.1$), the authors highlight that real-world haptic signals are highly correlated. In practical telepresence applications, the Weber Quantizer has demonstrated packet rate savings exceeding 90% without any perceptible degradation in the user's experience.
Conclusion and Future Work
The Weber Quantizer offers a robust framework for reducing network load in haptic systems by focusing on human perception rather than raw data precision. While the current analysis assumes a simple uniform distribution, it provides a foundation for more complex models. Future research will explore multidimensional signal processing (e.g., 3D force vectors) and the application of more sophisticated stochastic models, such as Wiener processes, to better characterize real-world haptic interactions.
Summary updated