We have now sufficient evidence that using electrical biosignals in the field of Alternative and Augmented Communication is feasible. Additionally, they are particularly suitable in the case of people with severe motor impairment, e.g. people with high-level spinal cord injury or with locked-up syndrome. Developing solutions for them implies that we find ways to use sensors that fit the user's needs and limitations, which in turn impacts the specifications of the system translating the user's intentions into commands. After devising solutions for a given user or profile, the system should be evaluated with an appropriate method, allowing a comparison with other solutions. This paper submits a review of the way three bioelectrical signals - electromyographic, electrooculographic and electroencephalographic - have been utilised in alternative communication with patients suffering severe motor restrictions. It also offers a comparative study of the various methods applied to measure the performance of AAC systems.

No abstract

No abstract

Patients with severe motor impairments, victims of stroke, amyotrophic lateral sclerosis and spinal cord injury are prevented from oral and gesture communication, demanding alternative channels and methods of communication, possibly using a computer. In order to obtain the complete emulation of a standard mouse, the single-click and double-click actions are desirable functionalities. In this study, the implementation of such actions is executed by the analysis of the electromyographic signal recorded from the Frontalis muscle. Muscle activity is discriminated from noise and this information is used to feed a state-machine that in turn decides which action is intended. The method uses an adaptive threshold, which offers freedom for the selection of the parameters of the system. The rate of successfully detected commands found was up to 100% for the single-click and 92% for the double-click. Even though good results were found for double-clicks, the experiment indicate muscle fatigue in the short term. The time response found was below 300 ms suggesting real-time implementation is feasible. Also, other devices can be operated with this approach, if it is accepted as a two symbols system generator.