@MP Divakar: Thanks for your comment! We at Hillcrest agree that in some applications it is most effective to include the fusion algorithms in the remote control. Our Freespace® MotionEngine™ software is modular so you can implement full processing on the remote or just sensor management on the remote with fusion, calibration, etc on the host (e.g. TV SoC). If full motion performance with the lowest power and cost is desired, then more processing on the host is generally a great solution. In the end though, the optimal choice depends on the particular application and customer design objectives.
@R. Colin Johnson: interesting article... Lucien's statement that "...have to transmit raw sensor data from the MEMS chips" passes the burden (not to mention process steps such as filtering and amplifying received signals) to the processors in the TV! I thought the most efficient use of gesture control data (accelerometer & gyro outputs) was thru fusion at the sensor interface itself.