The way people meet each other is usually face to face. Meanwhile, the way people maintain their contacts is mostly through social media. This results in a gap of translating a handshake into a digital connection. Shake-On is a start-up that has come up with an answer by means of a smart bracelet. Their aim is to wirelessly exchange contact details between users wearing the bracelet. This exchange is triggered by the most common human gesture people use when introducing themselves: the handshake. This thesis will overcome two major challenges for Shake-On. First, no general pattern recognition method can be applied to detect handshakes. This is caused by the fact that handshakes gestures show large variations among individual persons. Second, the system should be robust to multiple handshakes happening concurrently. This applies to the scenario of more than two people shaking hands while standing close to each other. Contact details should only be exchanged between people that are handshaking, which requires handshake matching. Again, large variations in `handshaking style' make it a cumbersome task to identify matching handshakes. This thesis proposes a two-fold solution to address the above-mentioned challenges. The first part includes handshake detection, using new features for pattern recognition that are tailored to handshaking recognizing. The second part proposes a new method to perform handshake matching that overcomes the shortcomings of existing solutions. The work done in this thesis has led to the following results: 1. The developed detection method takes into account limited resources and is therefore suitable for implementation on a smart bracelet. Moreover, it shows similar performance as the state-of-the-art solutions, namely an accuracy of 95%. In contrast to existing solutions using 6 stochastic features, our solution uses 4 computationally lightweight features. 2.Being the first of its kind, the matching method proposes a novel technique that maps handshakes to an abstract binary format. This format is called peakmaps. Because it eliminates personal handshaking style, peakmaps result in a pairing accuracy of 80% compared to 24% using basic cross correlation.