A method for detecting user head motion during an authentication transaction is provided that includes generating, by a processor, a motion type feature vector and a user head motion type prediction based on data generated for a sequence of frames. The frames are included in biometric data captured from a user. Moreover, the method includes generating a liveness rating feature vector based on the generated frame data, merging the motion type and liveness rating vectors, and generating a spoof prediction from the merged vector. When the generated spoof prediction indicates biometric data in the frames was spoofed, the method includes changing the user head motion type prediction to no motion. The method also includes storing the user head motion type prediction in a buffer and determining a final user head motion type detected for the frames.