Bent Schmidt-Nielsen

Evaluation of Different Speech and Touch Interfaces to In-Vehicle Music Retrieval Systems

In-vehicle music retrieval systems are becoming more and more popular. Previous studies have shown that they pose a real hazard to drivers when the interface is a tactile one which requires multiple entries and a combination of manual control and visual feedback. Voice interfaces exist as an alternative. Such interfaces can require either multiple or single conversational turns. In this study, each of 17 participants between the ages of 18 and 30 years old was asked to use three different music retrieval systems (one with a multiple entry touch interface, the iPod, one with a multiple turn voice interface, interface B, and one with a single turn voice interface, interface C) while driving through a virtual world. Measures of secondary task performance, eye behavior, vehicle control, and workload were recorded. When compared with the touch interface, the voice interfaces reduced the total time drivers spent with their eyes off the forward roadway, especially in prolonged glances, as well as both the total number of glances away from the forward roadway and the perceived workload. Furthermore, when compared with driving without a secondary task, both voice interfaces did not significantly impact hazard anticipation, the frequency of long glances away from the forward roadway, or vehicle control. The multiple turn voice interface (B) significantly increased both the time it took drivers to complete the task and the workload. The implications for interface design and safety are discussed.

A comparison between spoken queries and menu-based interfaces for in-car digital music selection

Distracted driving is a significant issue for our society today, and yet information technologies, including growing digital music collections, continue to be introduced into the automobile. This paper describes work concerning methods designed to lessen cognitive load and distracting visual demands on drivers as they go about the task of searching for and listening to digital music. The existing commercial paradigms for retrieval—graphical or spoken menu traversal, and text-based search—are unsatisfactory when cognitive resources are limited and keyboards are unavailable. We have previously proposed to use error-tolerant spoken queries [26] combined with direct modalities such as buttons mounted on the steering wheel [7]. In this paper, we present in detail the results of an experiment designed to compare the industry standard approach of hierarchical graphical menus to our approach. We found our proposed interface to be more efficient and less distracting in a simulated driving task.

An FFT-based companding front end for noise-robust automatic speech recognition

We describe an FFT-based companding algorithm for preprocessing speech before recognition. The algorithm mimics tone-to-tone suppression and masking in the auditory system to improve automatic speech recognition performance in noise. Moreover, it is also very computationally efficient and suited to digital implementations due to its use of the FFT. In an automotive digits recognition task with the CU-Move database recorded in real environmental noise, the algorithm improves the relative word error by 12.5% at -5 dB signal-to-noise ratio (SNR) and by 6.2% across all SNRs (-5 dB SNR to +5 dB SNR). In the Aurora-2 database recorded with artificially added noise in several environments, the algorithm improves the relative word error rate in almost all situations.

DT controls: adding identity to physical interfaces

In this paper, we show how traditional physical interface components such as switches, levers, knobs and touch screens can be easily modified to identify who is activating each control. This allows us to change the function per-formed by the control, and the sensory feedback provided by the control itself, dependent upon the user. An auditing function is also available that logs each users actions. We describe a number of example usage scenarios for our tech-nique, and present two sample implementations.

A speech-in list-out approach to spoken user interfaces

Spoken user interfaces are conventionally either dialogue-based or menu-based. In this paper we propose a third approach, in which the task of invoking responses from the system is treated as one of retrieval from the set of all possible responses. Unlike conventional spoken user interfaces that return a unique response to the user, the proposed interface returns a shortlist of possible responses, from which the user must make the final selection. We refer to such interfaces as Speech-In List-Out or SILO interfaces. Experiments show that SILO interfaces can be very effective, are highly robust to degraded speech recognition performance, and can impose significantly lower cognitive load on the user as compared to menu-based interfaces.

Evaluation of Two Types of In-Vehicle Music Retrieval and Navigation Systems:

It has been documented that use of an In-Vehicle Information System [IVIS] can impair driving performance. In general, voice-controlled IVISs are considered less distracting than those controlled by touch and are therefore considered as less interfering with the driving task. However, certain types of voice-controlled IVISs may be better (less distracting) than others. In this study, we evaluated two variants of a voice-controlled in-vehicle music retrieval and navigation system in order to investigate the effect of a common affordance where the IVIS controls the pace and timing of a multi-turn interaction. Participants were asked to drive various scenarios in a simulator while engaging with either a user-paced IVIS or a system-paced IVIS. Driving performance measures and IVIS usability measures were collected. Our results indicated that when the driving task imposes higher mental effort (e.g., on a curved roadway), the user-paced IVIS was found to be more effective and safer to use, whereas when the driving task imposes low mental effort (e.g., on a straight roadway), the user-paced and the system-paced IVIS was less effective but still safer.