Enabling Input on Tiny/Headless Systems Using Morse Code

Entering text on devices with tiny or no displays (referred to as headless devices) remains a challenge. Conventional virtual keyboards are ineffective on devices with tiny displays, like smartwatches or smart thermostats, since the keys of these keyboards are too small for precise selection (Arif and Mazalek, 2016). Entering text on a headless devices, such as a digital tangible in a tangible-tabletop system, is even more challenging due to the difficulties in verifying or correcting an input without visual feedback on a display. This work investigates the potential of Morse code as a method for text entry on these devices. Morse code encodes characters as standardized sequences of dots (.) and dashes ($-$) (Fig. 1). Since it was originally designed for telegraphs, a device without a display (Snodgrass and Camp, 1922), we hypothesize that it can enable users to enter text on both tiny and headless devices. In this work, however, we focus only on the learning of the code.

Not much work has investigated the potential of Morse code as an input method on computer systems. The earliest work in this area studied the effectiveness of Morse code as an input method with a Z-80 based microcomputer (Levine et al., 1986). In the study, six participants yielded a 7.9 wpm entry speed after two months of training (2–3 sessions per week). A different work investigated if Morse code can be learned through passive haptic learning (PHL) using the bone conduction transducer on a Google Glass (Seim et al., 2016). In their study, participants reached an 8 wpm entry speed after four hours of exposure to passive stimuli. A follow up study demonstrated PHL of Morse code on a smartwatch (Seim et al., 2018). Table 1 summarizes the findings of these studies.

Several works have also explored the potential of Morse code with people with physical disabilities (Koester and Arthanat, 2018) and with various channels, including eye-blink (Mukherjee and Chatterjee, 2015), foot stomping (Pedrosa and Pimentel, 2014), and tongue gestures (Sapaico and Sato, 2011). Some have used language models to improve text entry with Morse code (Tanaka-Ishii and Frank, 2005). These, however, are outside the scope of this work.

We developed a simple virtual keyboard based on Morse code. It enables users to enter characters using sequences of dots (.) and dashes ($-$). The keyboard has dedicated keys for dot, dash, backspace, and space (Fig. 2). To enter the letter “R”, represented by “$.-.$” in Morse code, the user presses the respective keys in that exact sequence, followed by the SEND key, which terminates the input sequence. The user presses the NEXT key to terminate the current phrase and progress to the next one. The keyboard does not use a predictive system, thus does not auto-complete words, auto-correct incorrect words, or suggest the next probable words.

We conducted a pilot study to investigate the learning of Morse code and the potential of Morse code as a method for text entry.

We only report descriptive statistics due to the small sample size ($N=2$) of the study.

Participants yielded a competitive entry speed with the Morse code keyboard (Table 1), reaching an average of 6.7 wpm by the final session. It is inspiring that learning occurred even in such brief sessions of the study. Entry speed improved by 16% in the second session and by 12% in the third session compared to the preceding sessions. The average entry speed over session correlated well ($R^{2}=0.9965$) with the power law of practice (Snoddy, 1926). This suggests that users are likely to get much faster with the method with practice. Participants made more errors in the first session (3.4%), which reduced by 76% in the second session. The error rate in the second and the final sessions were comparable (about 1%). Innately, the average error rate over session moderately correlated ($R^{2}=0.7158$) with the power law of practice (Snoddy, 1926). This suggests, unlike entry speed, error rate is unlikely to reduce substantially with practice. These findings indicate towards the possibility that Morse code could enable text entry on devices where using conventional input methods is impractical. Although it was not the focus of this work, Morse code could also enable people with various motor disabilities to enter text on computer systems.

There are several limitations of the work, which we will address in future studies. First, we evaluated the method on a smartphone where users could see their input. This visual feedback most likely improved their performance. It is unknown whether users will yield a similar performance with devices without a display. Second, we did not evaluate error correction in the study. Further investigation is needed to identify an effective method for identifying and correcting errors on devices without a display. Relevantly, studies showed that error correction is an integral part of text entry and difficulties in correcting errors significantly compromises users’ experience with a method (Arif and Stuerzlinger, 2010). Third, the custom keyboard includes dedicated keys for dots (.) and dashes ($-$), which may not be possible on all devices. Hence, investigation is needed to identify the best modes of interaction on devices that do not have physical or virtual keys. Some possibilities are using taps and long-taps, taps and hard-taps (de Jong et al., 2010; Heo and Lee, 2011; Arif et al., 2014; Zeleznik et al., 2001), or tap and flick (Baglioni et al., 2011) in place of dots and dashes. Finally, the study had a very small sample size ($N=2$). Replicating the study with a larger and more diverse sample will improve the generalizability of the findings.

We conducted a pilot study to investigate the potential of Morse code as a method for text entry on mobile devices. For this, we developed a simple Morse code keyboard. In the study, participants without prior experience with Morse code reached 6.7 wpm with the keyboard in three short sessions. Learning was observed both in terms of entry speed and accuracy, suggesting that the performance of the keyboard is likely to improve with practice.