Digital Signal Processing Reference
In-Depth Information
From these results (Table 2-1, Figure 2-5), we draw the following points:
1.
Employing delay-and-sum beamforming (DASB) or the proposed speech
adaptive beamforming (SA-BF), increases SegSNR slightly, but some
variability exists across speakers. These two methods are able to improve
WER for speech recognition by more than 19%.
2.
There is a measurable increase in SegSNR and a decrease in WER when
noise cancellation processing is activated (CSA-BF). With CSA-BF,
SegSNR improvement is +5.5dB on the average, and also provides a
relative WER improvement of 26%.
4.2
Environmental Sniffing
In this section we discuss our novel framework for extracting knowledge
concerning environmental noise from an input audio sequence and organizing
this knowledge for use by other speech systems. To date, most approaches
dealing with environmental noise in speech systems are based on assumptions
concerning the noise, or differences in collecting and training on a specific
noise condition, rather than exploring the nature of the noise. We are
interested in constructing a new speech framework which we have entitled
Environmental Sniffing
to detect, classify and track acoustic environmental
conditions in the car environment (Figure 2-6, see [24,32]). The first goal of
the framework is to seek out detailed information about the environmental
characteristics instead of just detecting environmental changes. The second
goal is to organize this knowledge in an effective manner to allow smart
decisions to direct other speech systems. Our framework uses a number of
speech processing modules including the Teager Energy Operator (TEO) and
a hybrid algorithm with segmentation, noise language modeling and
broad class monophone recognition in noise knowledge estimation. We
define a new information criterion,
Critical Performance Rate
(CPR), that
incorporates the impact of noise into Environmental Sniffing performance by
weighting the rate of each error type with a normalized cost function. We use
an in-vehicle speech and noise environment as a test platform for our
evaluations and investigate the integration of Environmental Sniffing into an
Automatic Speech Recognition (ASR) engine in this environment.
