Article Preview
TopIntroduction
ASD is seen primary stage and there are some deficiencies and it is not possible to identify behavioural patterns as symptoms unless it affects infant insignificant life. For the diagnosis of ASD, there are several tools and different methods taken by the therapist with different diagnostic tools. For the ASD predictions, many machine learning and data mining techniques were proposed but the prediction and accuracy fail when the population get increases. In existing projects, many researchers used brain signals for the ASD predictions, but the main drawback is more sensors and high signals will be used to take the signals.so it will be more complex and expansive apart from that it will create additional health issues to the patients. To overcome all this problem our research work concentrating on Model-free green environment to predict the autism spectrum disorder only by implanting the camera to take the various activities of a child or a person.
Generally, Gait is represented as the presentation of a person’s walking appearance. This walking style or appearance finds important in the biomedical fields to learn about the behavior of the individual. According to the persons walking style many other physical studies can be done and also helpful in surveillance applications. The parameters to be observed for the gait can either be gathered from using sensors or by using image processing tools (Pushpa et al, 2010) and (Pauline Luc, 2016).
Gait, in general, can be defined as how a person walks. An individual's gait is considered to be very important in the biomedical field as it provides valuable information regarding the individual normal and abnormal patterns, which can be used further in physical pathology and human surveillance.
There are two approaches to extract the quantitative parameters-either by the use of sensors, or by the help of Image Processing (Atiqur Rahman Ahad, 2013). The parameters in motion are collectively considered as temporal features. The temporal features such as joint angles can be observed using certain sensor which in turn is cost effective. Thus instead of learning using sensors, if the features are gathered from a recorded video of image sequence finds helpful to extract temporal information. Also real time video processing is also appreciated for gait analysis.
The movement of the person is identified and observed from the video frames and the features are classified using various tracking schemes. The recognition of motion related features are stamped as marker based and no- marker approach.
The no-marker schemes are non-accurate and approximation schemes of feature analysis from acquired data sequence. Few no- marker approaches rely on the real time camera data and their 3D structures. The data captured from only one view angle do not hold all required motion details because of the sagittal view (Deepjoy Das, Alok Chakrabarty,2015), (Roland Zügner,2018), (Ahmed Mahmoud Hamad,2011), (Mandeep Singh,2013) and (Minhua Zhang,2018). Such sagittal view-based information are studied using either model based or model free approach. The model-based schemes use various models to fit and extract the feature information (Chandra Prakash, 2016), (Ana Patrícia Rocha,2014) and (Daehee Kim,2009). This model based schemes always rely on high resolution data which is not always possible. The major advantage of such models are they are less prone to noises and other external artifacts such as clothing and other accessories.