The impact of the stimulation frequency on closed-loop control with electrotactile feedback

Liliana Paredes, Strahinja Došen, Frank Rattay, Bernhard Graimann, Dario FarinaJournal of NeuroEngineering and Rehabilitation 2015 · DOI

Abstract

Effective somatosensory feedback is a cornerstone for the functional integration of upper-limb prostheses. This study investigates how the frequency of electrotactile stimulation influences real-time closed-loop control. Results indicate that while sensory data (e.g., grasping force) can be sampled at relatively low rates (~5 Hz), it must be delivered to the user via high-frequency electrical pulses (>25 Hz) to maintain control performance. This suggests that the quality of tactile perception—specifically "perceptual fusion"—is more critical for effective human-in-the-loop control than the raw information transfer rate.

Introduction: The "Numb" Prosthesis Problem

Modern prosthetic limbs are sophisticated mechanical devices, yet they lack the intuitive sensory feedback inherent to biological limbs. Users often rely exclusively on visual cues to monitor their interactions with the environment, which is cognitively demanding and inefficient. Electrocutaneous stimulation offers a method to "write" sensory information back to the nervous system by applying electrical pulses to the skin.

A critical design question for these systems is determining the optimal stimulation frequency. In traditional control theory, higher frequencies equate to higher data bandwidth. However, in biological systems, frequency also dictates the quality of the sensation. Low frequencies may feel like distinct "taps," while higher frequencies fuse into a continuous "vibration." This study explores how this perceptual transition impacts a user's ability to perform real-time tracking tasks.

Experimental Methodology

To evaluate the impact of frequency, researchers recruited 10 healthy subjects to perform a compensatory tracking task. Subjects used a joystick to keep a cursor centered on a screen while it was perturbed by pseudorandom "wind" signals.

Experimental setup for real-time closed-loop control simulation.

The tracking error—the difference between the target and actual position—was relayed to the subject via electrotactile feedback. Two concentric electrodes were placed on the forearm: a dorsal channel for positive errors and a volar channel for negative errors. The study decoupled two key variables:

  1. Tracking Error Sampling Rate ($F_{TE}$): The frequency at which the system measured the error.
  2. Stimulation Frequency ($F_{STIM}$): The rate of electrical pulses delivered to the skin.

A "sample and hold" strategy was employed to isolate the effects of perception from data bandwidth. In this condition, the error was sampled slowly (5 Hz) but delivered at a high pulse rate (100 Hz) by repeating the sampled value. This allowed researchers to determine if performance degradation at low frequencies was due to a lack of data or poor perceptual quality.

Key Insights: Perception vs. Bandwidth

The study revealed that performance is primarily limited by the quality of tactile perception rather than the sampling rate of the underlying signal.

  • Perceptual Fusion: At low stimulation frequencies (5–10 Hz), subjects perceived discrete pulses, which were cognitively difficult to integrate into a control strategy. This led to "spatial confusion," where subjects struggled to identify which electrode was active, occasionally moving the joystick in the wrong direction.
  • The "Sample and Hold" Effect: When data sampled at 5 Hz was delivered at 100 Hz, performance almost entirely recovered to baseline levels. This demonstrates that the brain can effectively control a system even with low-rate data, provided that data is presented as a continuous, "fused" sensation.
  • Latency and Response Time: Lowering the stimulation frequency from 100 Hz to 5 Hz caused the user's response delay ($TD_{IO}$) to nearly triple, jumping from approximately 360 ms to over 1000 ms.

Results and Performance Metrics

The researchers used several metrics to quantify performance, including the Squared Pearson Correlation Coefficient (SPCC) and Normalized Root Mean Square Tracking Error (NRMSTE).

  • Baseline (100 Hz $F_{STIM}$ / 100 Hz $F_{TE}$): Subjects achieved high performance with an 87.5% correlation and a low 9.4% error rate.
  • Low Frequency (5 Hz $F_{STIM}$ / 5 Hz $F_{TE}$): Performance dropped significantly, with correlation falling to 61.1% and error increasing to 15.3%.
  • Optimized Feedback (100 Hz $F_{STIM}$ / 5 Hz $F_{TE}$): Using the sample-and-hold method, correlation returned to 83.6%, and the response delay dropped back to ~415 ms.

The findings suggest a critical threshold: stimulation should remain above 25 Hz to ensure the sensation is usable for real-time control.

Limitations and Future Directions

While promising, this study focused on able-bodied subjects and a 1D tracking task. Future research must validate these findings in amputee populations, where skin sensitivity on residual limbs may differ. Additionally, real-world prosthetic use involves multi-dimensional tasks (e.g., simultaneous grip and wrist rotation). It remains to be seen how the "fusion threshold" changes when multiple feedback channels are active, potentially increasing cognitive load. Long-term studies are also needed to assess sensory adaptation to high-frequency stimulation over time.

Summary updated

Abstract

The likely reason for the performance decrease and recovery was that the stimulation frequency critically influenced the tactile perception quality and thereby the effective rate of information transfer through the feedback channel. The outcome of this study can facilitate the selection of optimal system parameters for somatosensory feedback in upper limb prostheses. The results imply that the feedback variables (e.g., grasping force) should be transmitted at relatively high frequencies of stimulation (>25 Hz), but that they can be sampled at much lower rates (e.g., 5 Hz).
BibTeX
@article{paredes2015impact,
  author = {Liliana Paredes and Strahinja Došen and Frank Rattay and Bernhard Graimann and Dario Farina},
  title = {{The impact of the stimulation frequency on closed-loop control with electrotactile feedback}},
  journal = {Journal of NeuroEngineering and Rehabilitation},
  year = {2015},
  doi = {10.1186/s12984-015-0022-8},
}
Figure