Digital Signal Processing Reference
In-Depth Information
Figure 4-4.
Different configurations of microphone placement in the car environment: (a) A
microphone was placed in front of each speaker in the car; (b) Four microphones, linear
formation, were placed above the windshield of the car. The distance between two
microphones was 20 cm; (c) Four microphones, in square formation, were placed in the center
of the car ceiling. The side length of the square is 20 cm; (d) The microphones were attached
on the seat belts of each passenger.
We assessed the rate of correct active speaker determination. Each tester
uttered thirty short words for testing. Three conditions of car speed were
maintained in this experiment, i.e., idle, in-city, and on-highway conditions.
The idle condition means that the car was ignited but remained stationary. For
the in-city and on-highway conditions, respectively, the car speed was kept at
0~50 km/h and 70~100 km/h. Table 4-1 shows the speech identification
results of our experiments.
As shown in Table 4-1, the 4th configuration achieved the best
performance of ASD, since the vibration caused by the car's engine was
absorbed by the soft seatbelt. Configuration 3 was the worst because the
threshold was hard to determine (the difference between speakers was not
noticeable). For speakers in the front seats, configurations 1, 2 and 4 yielded
similar results. The reason was that under these configurations, the signals
from primary and secondary channels contained noticeable differences.
