Digital Signal Processing Reference
In-Depth Information
Table 10.1
Vehicles
considered for noise overlay
Vehicle
Derivative
Class
BMW 5 series
Touring
Executive car
BMW 6 series
Convertible
Executive car
BMW M5
Sedan
Exec. sports car
MINI Cooper
Convertible
Super-mini
Table 10.2
Road surfaces
and velocities considered for
noise overlay
Surface
Velocity (km/h)
Abbreviation
Big cobbles
30
COB
Smooth city road
50
CTY
Highway
120
HWY
would be mounted in the middle of the instrument panel. Consequently, noises as
occurring in the interior of a car have been recorded exactly at the same point.
The mouth-to-microphone transfer function had been neglected, since the masking
effect of background noise was proven to dominate over convolutional noise. As
interior noise masking varies depending on vehicle class and vehicle class derivates
[
21
], speech was superposed by noise of four different vehicles as they are listed in
Table
10.1
.
Besides the vehicle type, the road surface influences the characteristics of interior
noise. Hence, three different surfaces with typical velocities are further considered
as shown in Table
10.2
. A smooth city road at 50 km/h driving velocity and medium
revolution (CTY) provides the lowest excitation. At this profile noise caused by
wind, engine, wheels, etc. has its minimum. Higher excitation is measured for a
highway drive at 120 km/h (HWY), where wind noise is a multiple higher. The worst
case noise scenario is given by a road with big cobbles (COB). At 30 km/h, wind
noise can be neglected, but the rough cobble surface results in dominant wheel and
suspension noise. Figure
10.1
shows the SNR histograms of the accordingly noisy
speech utterances.
In spite of SNR levels below 0 dB, speech in the noisy test sequences is still
well audible since the recorded noise samples have most of their spectral energy
contained in the frequency band from 0 to 500 Hz (cf. Fig.
10.2
). As a result, there is
Fig. 10.1
SNR level his-
tograms for noisy speech
utterances [
10
]
