Article Preview
Top1. Introduction
Gait is ones walking pattern from which we can recognize a person. Gait recognition helps to recognize a person from behind, from far distance and in low light condition. C. D. Lim et al. (Lim et al., 2015a) Described in their paper that how gait analysis can improve gait of Parkinson Disease patient. In this work, they used depth camera attached with robotic walker to capture leg movement of patient. Parkinson Diseases effect on nervous system. In this disease, people lose their ability to move limbs in proper way and find difficulty in moving and walking. Symptoms continue and limbs become dysfunctional over time. Authors in this work, stored stride length and gait speed of few selected healthy older people whom they called control group and they selected few Parkinson Disease patient whom they called PD group. All people were male and aged between 69-92 years. They found that average stride length of control group is 37.2 cm which is greater than the stride length of PD group 22.38 cm but gait speed is almost equal in both cases. Early detection of Parkinson diseases (Kondragunta et al., 2019; Mallikarjuna et al.,2020) can delay the effect of this disease. For early detection of this disease deep learning algorithms can be used to estimate 2D poses. Then this 2D poses are projected in a 3D environment to extract gait parameters such as stride length, step angle. Deep learning is highly effective for 2D pose estimation. Mapping depth data with the 2D pose estimation produce 3D pose which helps to analysis gait for early detection. Rida et al. (Rida et al., 2015) showed that gait can be influenced by people clothing and carry bags which are termed as(Ali et al.,2011)covariates. They applied gait energy image to detect features. All feature vectors are processed with mask based variants. This work helps to identify gait in various clothing condition and was an improved system which elaborately explain ratio of the number of well classified samples over the total number of samples. Y. Qi et al. (Qi et al, 2016) describes gait phase detection from foot trajectory. They detected accurate gait cycle with error -0.02 ± 0.01 second and stance phase with error 0.04 ± 0.03 second and swing phase with error -0.05 ± 0.03 second. The aim of the work is to detect the influence of walking speed over gait spatio-temporal parameters. Gait analysis is widely used (Mastrigt et al., 2018) in forensic works. In several criminal trials in Europe, forensic gait analysis plays a major role in last 15 years. Clinical gait analysis and forensic gait analysis are not a completely different field of study because almost every feature is used in forensic gait analysis which are used in clinical gait analysis. These features include person’s posture, movement of hands, head tilt, distribution of weight, angle of knee joint during walking, strides length etc. Gait recognition is widely used in forensic works for its discriminating ability from person to person (Muramatsu et al., 2013). In his PhD thesis (Seely, 2010) Richard D. Seely described large multi-biometric dataset that was captured as video processing and later it was reconstructed as a 3D dataset. Introduction of prototype biometric tunnel and its evaluation through rigorous analysis of dataset and sensor modification gives satisfactory result to identify discriminating boundary of gait recognition. Gait recognition can be done (Sun et al.,2019) using deep learning to find the key points of skeleton from image. This helps to recognise weather the subject is walking or running. The key points of human skeleton are very important for human behaviour recognition and also posture detection. Event based (Wang et al., 2019) gait recognition, EV-Gait shows 96% recognition accuracy on DVS128 and EV-CASIA-B gait databases. It is considered that dynamic vision sensors (DVS) are new sensing modality for object recognition. It generates asynchronous noisy events. EV-Gait is able to work with this noisy event streams and showed a high accuracy level in gait recognition. Gait recognition (Sokolova andKonushin,2019) by neural features obtained through DVS has also shown high accuracy rate which is more than 98% for TUM-GAID gait data set. Cross-view gait (Chao et al.,2019) recognition form Gait Set shown 95.0% average rank-1 accuracy on CASIA-B dataset without covariates and this accuracy rate is 87.2% with bag-carrying and 70.4% with coat-wearing on CASIA-B data set. According to the psychological study (Boyd and Little, 2005) gait has small but very strong features for recognition of individual gait. There are several factors that can confuse the gait recognition system. These factors include variation of footwear, injury, change of age. Quasi gait recognition is very effective for gait recognition because it is less sensitive to these factors. Independent component analysis ICA (Lu and Zhang, 2007) and modified ICA (Rani and Arumugam, 2010) are also very effective for gait recognition. ICA is similar like Principal component analysis PCA where the difference is that ICA has statistically independent components. In research work on gait and posture (Papi et al., 2018) it is found that peak knee sagittal angels are plays an important role to diagnose movement disorder. Diseases like ataxia, restless legs syndrome, Parkinson's disease, and stroke are common sign movement disorder. In these cases, knee angles are interest of study of bio clinical laboratories. Recent development of wearable technologies (Margiotta et al., 2016; Papi et al., 2017; Duc et al., 2014) has made such measurement easy and accurate. Rehabilitation of stroke patients who have survived from trauma (Carbonaro et al., 2015) needs a daily life monitoring system. A textile goniometer work on knitted piezoresistive fabrics (KPF) developed to measure knee flexo-extension activities. This work (Chao et al., 2019) compares the performance of textile goniometer with inertial measurement unites(IMU). IMUs are the gold standard wearable motion sensors widely used. Goniometer gives same reading information of IMUs. This new light weighted goniometer gives extra set of information which IMUs does not. Ataxia (Putzhammer et al.,2005), schizophrenia are such diseases in which people lose his ability to think, feel, behave properly, has common motor disabling symptoms. Gait analysis for schizophrenia patient produces day to day mental ability improvement data for medical analysis. Anatomy of human nervous system (Sanders and Gillig,2010) reveals that walking messages are controlled by motor, premotor cortex, subcortical nuclei, brainstem and cerebellum. Motor and premotor cortex generates initial signal for walking and others are responsible for modifying the signal. Gait reflects every kind of nervous system condition. This paper gives a cause of different abnormal gait for different diseases and drug effect. Asymmetry stride length disturbed due to short in cortical and basal ganglia disorders. Chronic alcoholic patient and drug effect cause ataxic gait abnormality. Many works had been carried out on gait analysis for medical treatment.