A major obstacle to the implantation of User-Centered Design in the real world is the fact that no precise definition of the concept of usability exists that is widely accepted and applied in practice. Generally speaking, the literature tends to define usability in overly brief and ambiguous terms and to describe its application in informal terms. This is one of the main reasons why ad hoc techniques predominate in usability study methodologies. The aims of this article are to investigate the concept of usability and to describe it by means of a detailed taxonomy that is organized hierarchically and that contains exhaustive descriptions of usability attributes. This taxonomy can be used to support different stages in the development of usable systems.

Electroencephalographic arousals are defined as abrupt shifts in electroencephalogram (EEG) frequency during sleep. Occurrence of arousals results in fragmented sleep, being one of the most important causes of daytime sleepiness among sleep disorders. Detection of arousals requires a polysomnographic (PSG) recording to be made during the patient's sleep. The resulting PSG is then analyzed offline by the physician. This is a time-consuming task, hence, automation of this process is pursued. The analysis, which involves the correlation of various events in time occurring among the different channels, in conjunction with the complexity of the related biomedical signals, makes this task also difficult to achieve in the computer algorithm. In this paper, we present a method for the detection of EEG arousals working on multichannel PSGs. The algorithm detects arousals using the information available through two EEG channels and the electromyography. A signal-processing technique is first proposed for the analysis of biomedical signals and extraction of relevant information. Individual events are detected from the signals and subsequently are related in time. Finally, a classification phase carries out the final decision on the presence of the event. Classifiers based on Fisher's linear and quadratic discriminants, support vector machines and artificial neural networks are compared at this phase. Experiments conducted on 20 patients reported a sensitivity and specificity respectively of 0.86 and 0.76 in the detection of the arousal events.