3D pointing devices are indispensable in virtual reality (hereafter VR) and human-robot interaction scenarios. Existing devices are cumbersome or non-immersive or have a limited volume of operation. Hand gesture-based interfaces do not suffer from these problems and can be used for 3D pointing purposes. However, there is a lack of robust, accurate hand gesture-based pointing techniques which can be attributed to the non-existence of large and accurate data-set for the same. To overcome this barrier, we propose a data-set consisting of depth images with a large number (107000) of samples collected from 11 subjects, with accurate ground-truth and adequate variation in the orientation and distance of the hand w.r.t. the camera. We propose a 3D convolutional neural network based technique that works on the proposed data-set and achieves an accuracy of 94.49% for an angle error threshold of 10 degrees. The proposed data-set may be used for developing more accurate, robust, less computationally expensive methods.

Absolute orientation estimation is crucial for navigation of robots, drones, unmanned vehicles. It is also needed in bio-metrics, virtual reality, human robot interaction systems and in devices like cell phone, smart watch etc. Nine axis MEMS inertial motion unit(hereafter called IMU) consisting of accelerometer, gyroscope and magnetometer is used widely to obtain absolute orientation. Sensor fusion is used to combine the readings of the individual sensors to obtain correct estimate of the absolute orientation. However MEMS devices are noisy and correct calibration is needed for the sensor fusion to work. Existing libraries on nine axis sensor fusion fail to give accurate drift free orientation estimate as they either don't use accurate calibration algorithms or don't prescribe a way to collect good calibration data. Also it is difficult and time consuming for a hobbyist or a researcher working on bio-metrics or SLAM to implement complex calibration algorithms or to design good calibration rigs. We propose a new calibration setup which consists of some easy to implement calibration algorithms along with a new calibration rig. Our calibration framework attains better accuracy and drift free estimate of absolute orientation than the current state of art libraries.

Pointing direction estimation(PDE) is an important component of many Human Computer Interaction(HCI) applications such as human robot interaction, virtual reality, home automation, driver assistance systems etc. Existing vision based methods using only RGB data need multi-camera setup and are prone to problems due to variation in illumination and in skin color. Depth based techniques mostly use multiple body parts like head, shoulder etc. This imposes restriction on the user location and orientation w.r.t. camera. The only depth based method which uses only the hand region is restricted to a table top scenario and a small range of angles. To circumvent the above problems we propose a depth data based PDE technique that uses only the hand region and is better than existing techniques in terms of range of angles and robustness to variation of depth of the hand w.r.t. the camera while returning an accuracy that is comparable to existing techniques.