Fixation-identification in dynamic scenes: comparing an automated algorithm to manual coding