|
|
KEYWORDS: Non-speech audio, auditory icons, auditory illusions, feedback, notification.
Our approach structures the sound space using a set of organizing concepts useful to application designers. This approach builds on Gaver's auditory icons [4] and, to a lesser extent, on Blattner's earcons [2]. It contrasts with most uses of sound in today's user interfaces, which play pre-recorded audio clips. In order to make it easier for applications to include audio cues, we have developed the ENO audio system [1]. ENO is based on the concept of sound sources that produce sound when subjected to an interaction. For example, a machine source produces sound when it is started, and an object source produces sound when it is hit or scraped. The sounds are generated in real- time and can be controlled in real-time by high-level parameters such as machine speed or object material.
We believe that auditory illusions have an interesting potential for user interaction. Like visual illusions, they exploit imperfections in our perceptual systems. For example, a common visual illusion in computer graphics consists of generating an animation by displaying a series of fixed images. In this paper we introduce the use of an auditory illusion known as Sheppard-Risset tones to convey the idea of motion in the user interface. We first introduce Sheppard-Risset tones and then describe how they can be used for feedback when interacting with a scrollbar and for monitoring the progress of long system operations.
Figure 1: a semi-logarithmic plot of the spectrum of a
Sheppard-Risset tone. Over time, the partials' frequencies
are shifted upwards.
Sheppard-Risset tones are controlled by 4 parameters :
Scrollbars are used in graphical user interfaces to allow users to view large documents through small windows. A typical scrollbar has two arrows for incremental moves, a thumb (sometimes called an elevator) for direct access and two paging areas, between the thumb and the arrows, for incremental moves by pages.
The main problem with scrollbars stems from the constant shift in attention between two points: the document being scrolled, and the scrollbar itself. One specific problem, known as kangarooing [3], occurs when paging: when the thumb is close to the position being clicked in the paging area, successive clicks page up and down because the thumb jumps around the clicked position. In general, it takes a while for users to understand what is happening, especially since they are not watching the scrollbar. Another problem occurs when the cursor is inadvertently moved out of the scrollbar: users do not notice it and get confused when the interface does not respond as expected.
The elevator sound solves these problems: whenever the cursor is in the scrollbar, a faint sound is played. Its speed is zero in the thumb, slow in the arrows and faster in the paging areas. This way, users know where they are without looking. When clicking or dragging, the sound gets a bit louder. If kangarooing occurs, the user is immediately notified by the fact that the sound changes direction.
This design has a drawback: it does not provide audio feedback of the relative position in the document. We have tried to change the base frequency of the sound to reflect the position in the document, but this tends to conflict with the motion effect. In addition, since most listeners cannot recognize absolute pitch, it is not a good idea to use frequency to convey an absolute value. This is still an open problem.
The second use of the elevator sound is for audio progress bars. Progress bars act as a notification mechanism when the system is engaged in a long operation, such as copying a large file, and the progress of the operation (percent-done) is known. The problem with visual progress bars is that they need to be looked at to know the progress of the operation. Since the operation is long, the user is likely to engage in other tasks.
We have used the elevator sound to complement the visual display of a progress bar: the speed conveyed by the sound reflects the speed of the progression. Therefore, the sound conveys the instantaneous speed at which the operation progresses rather than the current "percent-done". This way, it is very easy to know when the operation stops progressing: the sound stops moving. This technique also works when the percent-done is not known but the rate of progression can be monitored. For example, when downloading a document, the Netscape World-Wide Web browser displays the current throughput of the connection, even if the size of the document is not known. The throughput can be mapped to the speed of the lift sound, giving an accurate and non-intrusive way to monitor the progress of the operation.