**4. Visual sensor networks**

*Data Acquisition - Recent Advances and Applications in Biomedical Engineering*

patients to be continuously monitored in all locations. In case of health emergency, an alert can be immediately generated to inform the medical staff that urgent intervention is necessary. The data collected by WBANs can also be used to localize person [3] and analyze movement of the body and recognize human activities. On this basis it is possible to develop systems that provide care and security for elderly persons. Moreover, the WBANs found applications in sports for performance monitoring of training activities, rehabilitation, disability assistance, and human–machine

*Wireless body area network (white circles depicts sensors, arrows correspond to data transfers).*

The above-discussed wireless sensor networks are usually built up with sensors that have the ability to sense physical parameters, perform basic processing tasks and transmit the collected data. More sophisticated solutions are equipped with smart sensors that have extended data processing capabilities. The smart sensors are capable of performing advanced data processing in order to make decisions and recognize relevant events [5]. This kind of sensors may use embedded machine learning algorithms to learn from collected data and to autonomously make assessments or predictions. In case of smart sensors, the data are processed locally. The sensor transmits results of information processing instead of the collected data. This approach leads to reduced data traffic, lower power consumption and latency,

An example of smart sensor is the solution discussed in [6] which uses a modified support-vector machine classifier for arrhythmia detection based on electrocardiogram signals and for seizure detection based on electroencephalogram signals. It was shown that the above-mentioned detection tasks can be performed

Another wearable smart sensor was proposed in [7] to detect and categorize cardiac arrhythmias from electrocardiogram readings. A convolutional-recurrent neural network was used in this solution. The neural network was adapted to perform the detection and classification tasks on embedded low-power processors

**2**

interfaces [4].

**Figure 1.**

**3. Smart sensors**

as well as to enhanced data privacy.

with a small memory footprint.

by low power wearable sensors in real time.

A special type of smart sensors are visual sensors, i.e., camera nodes equipped with embedded processor, and wireless communication module. The smart visual sensors have a number of potential applications, from security and patient monitoring to rehabilitation. For instance, in [8] a visual sensor was introduced for baby behavior monitoring in healthcare centers. This sensor detects abnormal motion of a baby and sends alerts to a user.

Visual sensors can be connected in visual sensor network (VSN). The camera nodes in VSN process image data locally, extract useful information, and exchange the information with other nodes. Using multiple camera nodes in the VSN provides different views of a monitored object, which improves the reliability of the recognized events [9].

In [10] a wireless VSN was proposed for supervision of patient rehabilitation. Results reported in the literature confirms that the VSN concept enables a low cost, light-weight and easy to use monitoring applications that meets tracking and localization needs of rehabilitation centers. An interesting example is the VSN, which was used to collect data for robot automation in rehabilitation of young children [11].
