TABLE 1.
Smart gait vocabulary.
| SG Vocabulary | Definition/Context in our papers |
|---|---|
| Artificial Intelligence | Artificial Intelligence is a technology that enables computers and devices to act intelligently and make decisions like humans (Amisha et al., 2019) |
| Machine Learning (ML) | Machine Learning is a subfield of AI that enables computers and devices to learn from data without being explicitly programmed (Mahesh, 2020). It includes supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. DL is a subfield of ML that extracts useful information directly from raw data to learn representations for pattern recognition (Esteva et al., 2019), (Pouyanfar et al., 2018). It is often referred to as the “black box” approach to reflect the abstract layers of human brain-like neural networks it consists of |
| Abnormal Gait Detection | The task of distinguishing a healthy gait from a pathological gait. Some of the pathologies that affect the walking pattern as discussed in this paper include dementia, Huntington’s disease (HD), PD, Autism Spectrum Disorder (ASD), Amyotrophic Lateral Sclerosis (ALS), Post-Stroke Hemiparetic (PSH), Acquired Brain Injury (ABI), depression, neuromuscular disease, lower extremity muscle fatigue, spastic diplegia, Cerebral Palsy (CP), etc. |
| Human Identification | Presented gait data is compared to a set of gait data with known identities (labeled training data) to determine whom the unknown gait belongs to |
| Human Re-identification | The task of identifying images of the same person from non-overlapping camera views at different times and locations. Gait is a behavioral biometric feature that is unobtrusive, hard to fake or conceal, and can be perceived from a distance without requiring the subject’s active collaboration (Nambiar et al., 2019) |
| Fall Detection | A binary classification task, usually concurrent with activity recognition that classifies an activity as fall or no fall |
| Activity Recognition | A classification task that maps features extracted from various sensor raw data to classes corresponding to activities such as sitting, lying, running, walking, stair climbing |
| Gender Recognition | Gender Recognition is a binary classification that maps features to qualitative outputs: male and female |
| Smart Home | A smart home utilizes context-aware and location-aware technologies to create intelligent automation and ubiquitous computing home environment for comfort, energy management, safety, and security (Hsu et al., 2017) |
| Gait Event Detection | Detection of a sequence of events that specifies the transition from one gait phase to another during each gait cycle. (Mannini et al., 2014) |
| Kinetic and Kinematic analysis | Kinematics studies the motion of body segments without considering masses or causal forces. Kinetics studies the relation between motion and its causes |
| Biometric Authentication | An automated method of verifying a person’s identity based on their biometric (gait) characteristics |
| Crowd Density | The density level of people in a crowded scene |
| Anomaly detection | It labels a behavior pattern that is "far away" from a trained model as anomalous, where “far away” is measured by a time-varying threshold (Sun et al., 2017) |
| Gait estimation from Pose | Parameters such as step length, stride length, stride time, cadence, etc., are estimated from the human pose |
| Human Gait Motion Modelling | A probabilistic manifold-based motion modeling framework able to model with a variety of walking styles from different individuals and with different strides (Ding and Fan, 2015) |
| Occupant Activity Sensing | Actively knowing the identity of the people within a monitored area and what they are doing (Yang et al., 2018) |
| Multi-Gait Recognition | Multi-gait is a term used by authors (Chen et al., 2018) to refer to the changed gait of a person walking with other people. Multi-gait recognition is the task of identifying a person when he is walking with different people |
| Brain-Computer Interface (BCI) | A technology that translates signals from human brain activity such as walking intention to a command sent to an external assistive, adaptive, or rehabilitative device, such as a prosthetic leg (Belkacem et al., 2020), (Khan et al., 2018) |
| Hybrid BCI (hBCI) | A system that fuses two bio-signals, where at least one is intentionally controlled. The different signals, such as data from electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS), are processed in real-time to establish and maintain communication between the brain and the computer. The output is evaluated through a feedback control loop. Compared with systems that use one modality alone, hBCI improves classification accuracy and the number of control commands by integrating the complementary properties of different modalities and removing artifacts (Khan et al., 2021) |
| Kinetic Energy Harvesting (KEH) | Technology that converts kinetic motion into energy. The individuality of the gait pattern can be captured in the output voltage signals of KEH systems, with the added benefit of energy savings, compared to accelerometers. Thus KEH systems are used as sensors and energy sources simultaneously (Lan et al., 2020) (Xu et al., 2021) |
| The Digital Human | Digital replicate of a human in the virtual space. Automatic, continuous gait monitoring will be an integral part of such systems (Zhang et al., 2020g) |
| IoT | IoT is a ubiquitous system of objects that are connected to the network, uniquely identifiable, capable of collecting, communicating, and processing data and AI-enabled to make autonomous decisions, individually or collectively (Chettri and Bera, 2019). Gait is an important biometric feature for continuous behavioral authentication in IoT systems (Liang et al., 2020) |