All data gathered by the BAN are personal/private data of the patient that need to be protected from non-authorised access for the complete data lifetime, as already mentioned. Therefore, data transmitted from sensor nodes to the base station (and to the PHR later on) need to be protected against eavesdropping and malicious damages or changes. The StrokeBack system makes use of the AES protocol, which provides a symmetric cipher for encryption of data streams. Since it is symmetric, both the sender and the receiver need to know the secret encryption key to be able to access encrypted data. In the simplest setting, this secret key has to be selected during the configuration phase of the sensor nodes and cannot be changed during usage of the system. This approach may be suitable for controlled trials, but will become more or less unsecure when the StrokeBack system would be used in its envisioned application for daily rehabilitation monitoring, since the key can then not be changed during operation. Thus, anybody who possesses the configured AES key can then in principle access all data gathered by the system, regardless whether this access is permitted.

To overcome these problems, the final BAN nodes are empowered by the IHP crypto-microcontroller that provides hardware for AES too, but additionally provides energy-efficient integrated hardware accelerators for the asymmetric cipher protocol ECC and the standardised Secure Hash Algorithm 1 (SHA-1) for generation of digital signatures and hashes. The latter can be used to detect unwanted or malicious changes of data, e.g. during wireless transmission. The hardware support for ECC (software variants for microcontrollers exist, but are much too heavy to be used on sensor nodes as side application) enables to implement a public–private key infrastructure in parallel even on the sensor nodes. This is used to implement access control directly on the sensor nodes and to enable secure exchange of new AES encryption keys at any time if needed. On the other hand, this ensures that no sender or receiver other than the authorised base station can communicate to the sensor nodes to exchange information or patient’s data.

Due to the non-precise and cheap clocks on the sensor nodes, these suffer from different clock drifts over time. That means, even if the clock of all sensor nodes is synchronised at the beginning of the day, e.g. during attachment, it may occur that the sensor data is not recorded exactly at the same time in parallel on all sensor nodes. Since sensor data is recorded at all nodes in intervals of 20 ms, the clock drift during the day must not exceed 10 ms in total to avoid incorrect correlation of motion data taken at different nodes. Communication facilities on the sensor nodes are to be used to synchronise all nodes and timestamps while recording data throughout the day.

For all communication issues between the sensor nodes and the base station mentioned before, each sensor node is equipped with a wireless communication module. According to previous estimations, it is assumed to transfer about 50 MB of data in total for each sensor node in use per day, including overhead data required for instance for packet headers, checksums, MAC (Medium Access Control) and routing data. Provided that a patient wears three sensor nodes in parallel for the maximum setting (wrist, upper arm, chest), about 150 MB of data have to be transferred to the base station during the recharging phase (at night). Taking the minimum time slot of 5 h for charging and transmission of all nodes into account, the BAN needs to transfer about 30 MB per hour. This equals to 500 Kbyte per minute or about 8.3 Kbyte per second, respectively. To summarise, to comply with the requirements set above, the communication module of each sensor node should provide a minimum transfer rate of 10 Kbyte per second (80 Kbit/s).

For wireless communication, the widely used Bluetooth (BT) standard has been investigated while focusing on using the more advanced BT-LE. Both would provide more than sufficient data rates to fulfil estimated transfer rates. The BT-LE allows data rates up to 1000 Kbit per second (1 Mbit/s). The BT standard even provides data rates of up to 2.1 Mbit/s (Enhanced Data Rate (EDR) mode), but usually consumes much more energy. Unfortunately, a dual BT front end that is compatible to both BT standards in parallel (called ‘BT Smart Ready’) was not available at the time of the design phase of the sensor nodes. Hence, a single BT-LE module was integrated.

Estimated raw sensor data of course need to be stored at the node throughout the day. Flash is the currently most widely used technology of non-volatile memory for small devices. In principle, two options of flash memory are available for our application, i.e. tiny flash modules directly assembled on the sensor board or flash memory cards that can be exchanged. Both options are relatively cheap, but feature different advantages and drawback.

Flash memory cards, or micro-SD cards in our case, are very cheap and available as mass product with capacities ranging from several hundred megabytes up to 64 Gigabyte. Further advantages are that they in principle are compatible to other devices and can be replaced easily and hence provide more flexibility. Their drawback is the required size for the SD-card slot on the board and their power consumption. Very small card slots still require a board area of about 12 × 12 mm, which is already half of the envisioned board size (refer to Sect. 3.9.1). The power supply for micro-SD cards requires around 3.0–3.3 V.

Flash modules for on-board assembly are slightly more expensive than SD-cards and provide less memory capacities. Nevertheless, such modules feature much smaller dimensions than a micro-SD slot and more important: the power consumption is significantly lower. Both features are of utmost importance for the StrokeBack sensor nodes. Hence, we decided to assemble a flash module with 128 MB storing capacity.

Energy efficiency is one of the key aspects when it comes to design of sensor nodes. The more energy the sensor nodes consumes during operation, the more energy need to be available and stored on the nodes to reach a certain minimum operation time. The minimum operation time for each node is assumed 12 h, as introduced before. However, not only the amount of energy consumed by the node itself need to be minimised of course. Obviously, using larger battery packs would solve most of energy supply problems, but there are several aspects that need to be considered as well, in particular from the usability point of view. Larger battery packs might not fit into very small and lightweight sensor cases and will unnecessarily increase weight of the sensor node.

Therefore, the StrokeBack BAN is supplied by very small rechargeable Lithium-Polymer batteries providing several hundred milliamps per hour (mAh). The sensors can commonly be charged via a standard micro-USB adapter. In addition, the StrokeBack sensor nodes integrate wireless power supply based on the Qi wireless power standard [36] for automatically recharging batteries when the sensor node is in charging position. This will enable getting rid of cables and cable plugs and therefore significantly improve patient’s experiences and acceptance.

The introduced monitoring cycle does not need to facilitate on-BAN movement classification by default. However, the ultimate research challenge of monitoring with body-worn sensors is enabling the BAN to classify and analyse motion data directly on the sensor nodes, without the need to use a base station for post-processing of sensed raw data. Therefore, the StrokeBack team investigated integration of a novel hardware module, called Application Specific Integrated Circuit (ASIC), for accelerating on-BAN movement classification while reducing the amount of recorded data at the same time. This would provide a novel and unique application.

Safety in this context means the BAN may not put the patients’ health at risk. The StrokeBack BAN is used for monitoring issues only and hence, it does not provide any actuators. Two more features may be considered, i.e. wireless communication and power supply. The exposure of wireless transmission should be kept minimal by reducing the communication frequency and transmission power. This can be neglected for the StrokeBack application scenario, since usage of radio transmission is almost reduced to recharging time when the patient does not wear the BAN. The power supply on the node is very low and the amount of energy stored is small. This should not be a risk for the patient. Furthermore, the charging module on the board includes autonomous temperature control and shuts down the sensor node in case of difficulty while charging and overheating problems.

A full list of chosen main components can be found in Table 3.1. Most of these components have been used on both platforms. However, differences in the set of integrated components between both platforms are marked.