Tactile perception and the optimisation of information transfer through the skin

Summary of work funded by EPSRC grants GR/J30820 & GR/J63163. Principal investigators: Dr Ian R Summers and Dr John C Stevens.

This project is a collaboration between the School of Physics, University of Exeter and the School of Medical Physics and Clinical Engineering, University of Sheffield. Measurements have been made on the perception of time-varying vibratory stimuli on the skin, with a view to establishing the optimum strategy for transmission of information via a single tactile channel. The results are particularly intended for application in the area of tactile aids for the hearing impaired, but they also have considerable interest in relation to other sensory-substitution systems and in relation to the study of tactile perception mechanisms in general.

In a first experiment, subjects were required to identify a one-octave step change in frequency presented to the index fingertip at various rates. Tests were carried out with sinewave stimuli and with pulse stimuli. Subjects performed significantly better with stimulation at 24 dB re threshold than at 36 dB re threshold, casting doubt on anecdotal evidence that strong stimuli are better perceived than weaker stimuli.

In a second experiment, subjects were required to discriminate between short sequences of stimulus elements in an "odd-one-out from three" task. Each sequence was presented as a frequency-and-amplitude-modulated stimulus to the index fingertip. Tests were again carried out with sinewave stimuli and with pulse stimuli. It proved harder for subjects to perceive gross features in a rapid sequence of elements than fine detail in a slow sequence of elements - sequences of six 80 ms elements in three possible categories were found to be less effective for information transmission than sequences of three 160 ms elements in nine possible categories, although the potential information transfer is the same in each case.

Data from the first experiment suggest a small advantage for sinewave stimuli over pulse stimuli, but no difference between the two is observed in the data from the second experiment. It can thus be concluded that pulse stimulation, which is much easier to achieve in practical devices with limited-bandwidth transducers, may be used with little or no detriment to performance.

In a third experiment, using pulse stimulation only, discrimination of sequences at the fingertip was compared to discrimination at the wrist (glabrous skin on the underside). Stimulation levels at the wrist were 10 dB higher than those at the fingertip, but stimuli were otherwise identical. Perception was significantly better on the wrist. This suggests that, contrary to the received wisdom from a variety of anecdotal sources, fingertip data in the literature may not necessarily represent an overestimate of the potential for practical devices using stimulation at other sites.

Data from the first experiment suggest that temporal features in isolated stimuli can be perceived down to a timescale of 50 ms or so, but results from the second and third experiments suggest that discrimination of sequences becomes difficult when the duration of stimulus elements is reduced to around 100 ms. This observation, together with an information transfer rate of 6 bits per second which was estimated for the 160 ms elements in the second experiment, suggests that frequency-and-amplitude-modulated tactile stimuli may be more suitable for transmitting suprasegmental speech information rather than the more rapid segmental information. This conjecture has been confirmed by direct measurements on the tactile transmission of these two types signal. However, it appears that perception of some segmental features is possible on the basis of gaps ("silences") in the speech-derived stimulus.

On the basis of these findings, a prototype aid has been constructed, miniaturised into an entirely wrist-worn format by use of a RISC processor for signal processing.

This work was funded by the UK Engineering and Physical Sciences Research Council.