Abstract
This study represents the first systematic evaluation of a Time-Division Multiplexing (TDM) based approach for closing the loop using electrotactile feedback in myoelectric systems. The overall conclusion is that TDM is a feasible and attractive method for closed-loop myocontrol, since it is easy to implement as a software-only solution, has limited impact on performance when using proper feedback window (FW) durations, and may decrease habituation due to the burst-like nature of stimulation delivery.
Introduction
Modern myoelectric prostheses allow users to control robotic limbs using electrical signals from their muscles (EMG). However, a significant limitation of current commercial devices is the lack of sensory feedback. Without tactile information, users must rely entirely on visual cues to determine if they are gripping an object with the correct force.
While electrotactile stimulation—applying small electrical pulses to the skin—is a promising method for providing this feedback, it presents a major technical challenge: stimulation pulses are significantly stronger than the tiny EMG signals used for control. When performed simultaneously, the feedback "noise" effectively drowns out the control "signal." While hardware-based solutions like blanking circuits exist, this research investigates a simpler, software-based Time-Division Multiplexing (TDM) approach that rapidly switches between recording and stimulation modes.
Experimental Methodology
The researchers implemented a closed-loop compensatory tracking task involving nine healthy able-bodied subjects (average age 29). Subjects were required to control a virtual prosthesis aperture to follow a randomized reference trajectory over 80-second sequences.
The core of the study involved varying the Feedback Window (FW)—the specific duration during which the stimulator is active and EMG recording is paused. Three TDM conditions were tested:
- TDM40: 40ms feedback window.
- TDM100: 100ms feedback window.
- TDM300: 300ms feedback window.
These conditions were compared against a "gold standard" benchmark: CONT-CLT (Continuous recording with Contralateral feedback), where recording and stimulation occurred on different arms to physically eliminate interference.
The TDM Mechanism and the "Sweet Spot"
TDM creates a fundamental trade-off between machine control and human perception. A shorter window (e.g., 40ms) allows for more frequent controller updates, which is technically better for the machine, but the human brain struggles to perceive such brief tactile bursts. Conversely, a longer window (e.g., 300ms) provides clear sensation but introduces perceived "lag" in the prosthesis because the control signal is updated less frequently.
The study identified 100ms as the optimal "sweet spot." This duration provides enough pulses for the user to reliably perceive the error signal while maintaining a command update rate that feels responsive and fluid.
Key Results and Performance Metrics
The evaluation showed that TDM is a highly effective alternative to complex hardware filtering:
- High Fidelity: Both TDM100 and TDM300 achieved a cross-correlation coefficient of approximately 0.95, performing remarkably close to the continuous benchmark.
- Perceptual Thresholds: TDM40 performed significantly worse across all metrics. Users reported difficulty "feeling" the signal, leading to increased delays as they waited for cumulative bursts to confirm their perception.
- Functional Accuracy: In the TDM100 condition, the average time delay increased by only 20% compared to continuous control. Furthermore, the tracking error (RMSTE) remained below 10% of the total trajectory range, proving that the intermittent "blind spots" in recording do not break the control loop.
- Reduced Habituation: An unexpected benefit of TDM is that the intermittent, burst-like stimulation may reduce sensory habituation—the tendency for the brain to ignore constant tactile stimuli over time.
Conclusion and Future Directions
This research demonstrates that TDM is a viable, software-only solution for closing the loop in myoelectric prostheses. Its portability makes it an attractive option for existing hardware platforms. However, several areas for future work remain. The study was conducted on able-bodied subjects; further research is needed to determine if amputees, who may have different skin sensitivity at the stump, require different optimal window durations. Additionally, moving beyond 1D virtual tasks to multi-axis physical robotic control will be essential to fully validate the approach in real-world scenarios. Future investigations may also explore "spatial coding," using TDM to deliver complex feedback across multiple electrode sites.
Summary updated