The use of electrical biosignals, like electroencephalogram (EEG), electromyogram (EMG) and electroneurogram (ENG), gain a lot of importance for the assessment of functions in the human body. These signals are used as major indicators for medical professionals, patients or professional athletes during diagnostic and monitoring processes. Furthermore, the biosignal-based intelligent control of prostheses or handicapped limbs is a key challenge in medical technology. In particular EMG and ENG are used to get information about the peripheral nervous system including information transfer due to sense data and motion control by peripheral nerves. Based on these signals a multitude of applications are existing or in the future envisaged; they range from the achievement of therapeutic objectives up to prosthesis control, for example, to operate an artificial hand or forearm. There are several requirements on a system existing to realize these functionalities:
Data acquisition and stimulation
The EEG, EMG or ENG data has to be acquired and sampled according their signal characteristics, given in Table 1. In particular, additional applications, like a stimulation, are necessary, for example providing the measurement of the nerve conduction velocity.
Data processing
The acquired data (action potentials) are disturbed by intrinsic noise. In addition, they are overlaid by a substantial extrinsic noise, originated for example by EMG from surrounding muscles. Therefore, we have to filter the recorded data with integrated analog filter and additional digital filter. There are several specific high-pass, low-pass, band-pass and notch filters available. A further data processing is necessary, on the one hand to improve the data condition due to asynchronous and aperiodic samples, and on the other hand to generate events from the action potentials like the activity level of a muscle group or the detection of an exposure scenario.
Identification
The identification feature is required for prosthesis control or any type of high level signal evaluation and correlation. The identification is based on machine learning and recognizes movement commands and inherent feedback signals (Verdult, 2002), (Wodlinger and Durand, 2011). The identification method and the corresponding verification scenario have been introduced in (Klinger and Klauke, 2013), (Klinger, 2014) based on results in [Bohlmann et al., 2010], [Bohlmann et al., 2011]. In this paper, we focus on the closed-loop verification approach.
Data archiving
After data acquisition and data processing the results have to be saved locally if there is no direct data transmission for an evaluation possible or if local data are required due to an offline analysis. Furthermore, for identification a certain data amount is necessary during the operating phase (Klinger and Klauke, 2013).
Data interfacing
Data has to be transmitted for evaluation or monitoring purposes to a host system.
User Interfacing
To select and execute certain functionalities and for online information an user interface must be available.
Configuration
Due to the different application scenarios and system functions a configuration is necessary.