Earlier we had problem sensing the values from knock sensor correctly. Vibrations transmitted along the medium to which the sensors were attached (white board of lecture hall in BIT) were making it difficult for us to accurately determine the sensor which was hit by the ball. The estimation was 70-85% correct. However, we needed more accurate results for the game, since it would not be a fun game if an accurate hit is treated as a miss. We needed more than 95% accuracy. This week we changed the placement strategy of sensors and provided enough insulation among sensors. A thick layer of tissue papers was placed between sensor and the white board to absorb the transmission of vibrations from one sensor to another. This decoupling worked well and accurate estimation of hits became possible.
Moreover, we determined the criteria of exactly detecting the knock on the sensor. If the hit is in diameter of about 15 cm around the sensor, the values from other sensors are exactly 0. If hit is made somewhere between two sensors (outside this diameter), the nearest sensor gets highest value. The other neighboring sensors receive a value normally less than 10. This helped us in determining the hit accurately to about 98% of the time.
The figure shows a screen-shot depicting four sensors attached to the artificial wall.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment