Text this: Recombining Features of Frequency Domain and Location for Machine Recognition of Sign Language