IEEE Consumer Electronics Magazine - January/February 2023 - 81

front-endprocessing framework, in particular tostress emotiondetectioncases (suchas
anger, sad, fear, andhappy) indifferent nonstationarynoisy environments, suchas car,
airport, traffic, and train. Thisarticle encompasses three interrelated issues:1) analysis,
modification, andsynthesis of noisy speechemotion inmodulationdomain in real-time
backgroundnoise, 2) extracting set of Mel-frequencycepstral coefficients features from
noisy speech stimuli for speechemotion recognition, and 3) evaluation of overall system
performancebymeans of objective parameters, andconfusion matrix inadverse
environmentsusing speechemotion database InteractiveEmotional DyadicMotionCapture.
Theexperimental results showthat favorableperformance in state-of-the-art stress
monitoring yields high levels of consumer satisfaction for security invehiclecomparison to
traditional frameworks.
& THE AUTOMATIC SPEECH emotion recognition is
a very recent topic of research in the human-
machine interaction field.
It has a wide range of applications such as incar
devices where mental state information of
the car driver is used to initiate their safety.1,2 In
automatic remote call center, it is used to timely
detect customers dissatisfaction. With the rapid
increase in the use of smart speech technologies,
the human beings can communicate with
the machines, which is used in automatic cars,
education, military, and medical applications .
Recent trend in Internet of Vehicle Things
(IoVT) promotes robust interactions in between
humans and vehicles, which ultimately points to
enhance human abilities, such as hearing, visual
surveillance, or emotion awareness, as a part of
safety concern.3
Figure 1 shows the functional block architecture
of the IoV system and the speech server.
The devices within client environment connect
with speech servers, IoT servers, and IoV servers
with various network layers and protocols.
The main contribution of this article is to
highlight the vital potential of speech methodologies
for improving the state of the art in IoVT,
which are listed as follows.
1) Preprocessing Methodology: To monitor the
human's emotion, preprocessing of real-time
voice captured by speech sensors is essential.
However, the noisy environmental conditions
hinder the IoVT system from this task.
Therefore, we propose a method that incorporates
various noise estimation techniques.
2) To reduce complexity: The computational
load introduces challenges in the implementation
of preprocessing methods; thus, we
need quick response in automated vehicles.
Therefore, the solution to conventional
methods is provided to reduce the computational
load along with a set of parameters.
3) To ensure security and quality of service (QoS):
In this article, we study the benefits of the proposed
optimized modulation spectral subtraction
(OMSS) speech enhancement for a speech
recognition system, in particular, the influence
of different nonstationary noises (background
environment such as airport, car, train, and
traffic) on typical speech emotion recognition.
Figure 1. Functional block architecture of IOV
system and speech servers.
4) Validation: In order to find the performance of
speech emotion recognition on the proposed
OMSS method, results were compared to the
traditional ModSpecSub method.4 The speech
emotion database Interactive Emotional
Dyadic Motion Capture (IEMOCAP) with various
SNRs is applied for the experimental
January/February 2023
81
IEEE Consumer Electronics Magazine - January/February 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - January/February 2023
