However additional sensors significantly increase the cost of the system. By utilizing data available from smartphones, the comparable cost is significantly reduced, only requiring users to download an appropriate software application and a cheap garment for securing their smartphone.Accelerometers have been used for human activity recognition in a large amount of existing work [10�C12]. Research has shown that accelerometers can be used to identify human activity for high energy actions such as walking, jogging, jumping, etc. [13]. In sports, accelerometers have been used to monitor elite athletes in competition or training environments. In swimming applications, accelerometers have allowed the comparison of stroke characteristics for a variety of training strokes and therefore have helped to improve swimming technique [14].

When used in competitive rowing and coupled with other monitoring techniques such as impeller velocity, they allow for the recovery of intra- and inter-stroke phases as a means to assess performance and this has been used by competition rowers to improve performance at national and international competitions [10].Most approaches in human activity recognition have relied on multiple expensive sensors. With the increase in smartphone ownership there has been more research conducted utilizing the sensors embedded within smartphones. Human activity recognition using smartphones have been employed to support patient monitoring [15], to identify the user’s current mobility [16] and for monitoring daily activities [17].

However in this work we will show how smartphones can be used to recognize human activity in sport. To the best of our knowledge this is the first such study conducted.In any activity recognition problem, feature extraction is a vital operation to determine those features with relatively small intra-class yet large inter-class variations that can be used as the basis for effective classification. It is preferable to have a low number of features due to the associated reduction of the computational load of the classification process. One method to extract discriminative features from a signal is to use the wavelet GSK-3 transform. The wavelet transform splits a signal into different frequency components, and then analyses each component with a resolution matched to its scale.

Wavelets have advantages over traditional Fourier methods in analyzing physical situations where the signal contains discontinuities and sharp spikes.In this paper, we take advantage of the embedded accelerometer within a smartphone and position it on the upper crevice of a user’s back as seen in Figure 1 to classify sporting events. There is a large amount of literature for activity recognition but it is limited for classifying sporting activities.