![](https://csdnimg.cn/release/download_crawler_static/10262109/bg13.jpg)
of digital noise, and automatic segmentation and recognition of the participants’
faces. In experimental section, one can find a digital comparison of three basic
techniques for face recognition, such as the principal components analysis, linear
discriminant analysis, and local binary pattern analysis. The authors developed a
method of audiovisual recording of participant activity based on the multifunction
system of video monitoring and multichannel system of audio localization that
detects the position of the sound source evaluating the phase difference of signals
recorded by pairs of microphones from four arrays [13].
Chapter 7 contributes in the audience analys is systems [14] and gender recog-
nition systems [15] using the assessment of face image quality. This is a continu-
ation of investigations of these authors during last two decades. Several standards,
such as ISO/IEC 19794-5 and ICAO 9303, contain a description of parameters that
provide a decision of the image suitability in the automatic face recognition sys-
tems. All parameters are grouped into two classes the textural features (sharpness,
contrast and light intensity, compression ratio, and other distortions) and face
features (symmetry, position, rotation, eyes visibility, and the presence of glare or
shadows on the face). For such algorithms, an assessment of face image quality is a
cornerstone because a success of face verification/identification depends strongly
from the selected lucky frames. Three types of the quality metrics were considered,
such as the texture-based, symmetry-based, and universal metrics. The authors
developed the no-reference image quality assessment algorithms based on the
analysis of texture and the landmark points ’ symmetry in the facial images [16].
Note the original methodology of the experiments’ statement. Two sets were
formed called as Top1 with the single best image of a person and Top3 considering
three high quality images chosen by the experts and objective metrics. The pro-
posed algorithms were tested using standard LIVE and TID2013 image database.
The performance results show that the proposed algorithms are highly competitive
for audience analysis systems and, moreover, have very low computational com-
plexity making it well suited for real-time applications. Additionally, novel gender
classification algorithm was suggested based on non-linear SVM classifier. It
includes several steps, among which are the scale-invariant feature transform,
histogram of oriented gradients, Gabor filters, and pre-selection of blocks with the
corresponding positive experimental results.
Chapter 8 contains a description of real time eye blink detection method for
Android device controlling and its implementation using standard libraries like
OpenCV conversion procedures. The Human Computer Interaction (HCI) or
Mobile Computer Interaction (MCI) may be realized in different ways. The authors
chose the issue of the eye blinking detection that has a significant role in the human
mobile interaction, computer interaction, driving safety, and health care. The
heuristic propositions concluded from the theory and practice of the HCI were used
during design. The eye tracking is the fastest non-invasive method of measuring
user attention [17]. The best definition of eye tracking can be the estimation of user
gaze’s direction, which is very difficult problem especially in the real world. The
proposed method consists of five main parts including the mobile camera pro-
cessing, face detection, eye detection, eye tracking, and blink detection. At the
1 Innovative Algorithms in Computer Vision 5