The CDC estimates that millions of adults aged 65 and older fall each year. In fact, more than one in four older adults fall each year, but fewer than half tell their doctor. 

These falls could be potential causes of serious injury among the elderly, especially visually impaired individuals, if they do not get timely medical help. While support may be near, lack of awareness or delays in alerting caregivers, volunteers, or family often prevent prompt assistance, causing severe consequences. 

To counter this, Sightica Innovations developed advanced AI-based monitoring to ensure constant safety and rapid response. Traditional single-sensor fall detection struggles in low light, noise, or clutter, but Sightica’s multi-modal fusion overcomes these gaps.

A high-accuracy AI solution

The Fall Detection Alert Device developed by Sunil J Mathew of Sightica Innovations is an affordable, high-accuracy AI solution that enhances the safety, confidence, and independence of elderly and visually impaired individuals. 

The initial prototypes are currently being tested in three elderly users’ homes, with plans to scale the solution to elderly care homes and support individuals with dementia, Alzheimer’s disease, and other care needs. 

The system uses AI-enabled RGB and infrared cameras for real-time, non-intrusive monitoring, leveraging the MoveNet Pose Estimation algorithm and OpenCV to detect and track users, extracting 17 body keypoints, including coordinates and confidence scores, to assess posture and movement. 

To protect privacy, the technology applies skeleton-based masking to video frames while enabling instant SMS and WhatsApp alerts, along with snapshots, to family members, caregivers, and healthcare providers through a mobile app. 

By combining privacy-first hardware and software with ongoing improvements through validated datasets, pilot deployments, and real-world feedback, the solution aims to enable rapid emergency response during the critical ‘golden hour’ following a fall, potentially saving lives and improving access to reliable care. 

Leveraging Artificial Intelligence and computer vision

The innovation integrates pose estimation, audio intelligence, and edge AI to address delayed fall detection and emergency response among elderly individuals. By using parallel video and audio intelligence pipelines to continuously monitor the environment, it improves detection reliability. 

The video system uses AI to detect people, track their body movements and posture, and check for sudden movements that may indicate a fall. At the same time, the audio system analyses sounds using different techniques to identify falls, screams, and normal sounds. 

To support monitoring in low-light and complete-darkness conditions, the innovation incorporates adaptive infrared (IR) illumination. High-confidence acoustic events can activate IR LEDs, restoring visibility for the visual pipeline and enabling further verification. Through Raspberry Pi devices, the data can be processed instantly on-site, while protecting privacy by showing body outlines instead of clear images and masking video footage. 

Under controlled testing, the integrated multimodal system has achieved approximately 95.45% fall-detection accuracy; the video-based pose classification subsystem achieves approximately 94% accuracy, while the audio classification subsystem achieves approximately 92% accuracy.

Why are falls tough to detect?

Across the deployments they have experienced, Sightica has realised that the primary challenges have been ensuring reliable fall detection across diverse environmental and behavioural conditions while minimising false positives. 

Everyday activities, such as sitting, sleeping, or resting, can resemble fall-related postures, while changing lighting conditions can affect visual detection. Sightica addresses these challenges by combining multiple AI techniques with audio intelligence and adaptive IR illumination, rather than relying on a single detection method. Edge-based processing, skeletal overlays, and privacy masking help reduce unnecessary exposure of visual information. Further real-world validation is essential to establish reliability across different environments before wider deployment.

The future scope of the innovation

The target audience for the Sightica innovation is visually challenged individuals, the elderly and those living alone. 

If scaled, it could prove to be a boon for elderly-care homes, individuals living with dementia or Alzheimer’s disease, and hostels for people with visual impairments. It could also benefit elderly individuals living alone by providing timely alerts to family members and caregivers in the event of a fall or emergency. By enabling quicker responses, the technology can improve safety, promote independence, and provide greater peace of mind to individuals and their families. 

Another boon is the nature of the innovation being a non-wearable monitoring system. 

A wearable solution could be forgotten, removed, or left uncharged, whereas the non-wearable monitoring system can operate in the background while the beneficiary follows their normal routine. 

From the family and caregiver perspective, the main value is the ability to receive an alert when a significant event is detected, even when they are not physically present. Sightica’s fall detection device therefore provides an additional layer of awareness between the elderly person and their remote caregivers. 

Why a multimodal AI approach is beneficial

Sightica Pro uses a combination of video and audio AI to detect falls more reliably, rather than relying on a single source of information. The video system detects people, tracks their body movements and posture, and checks for sudden movements that may indicate a fall. At the same time, the audio system identifies sounds such as falls and screams. By combining these two sources, the system can better distinguish falls from normal daily activities, even when visibility is poor.

When a potential fall or emergency is detected, the system can send alerts to designated family members or caregivers. This is especially useful for families who live away from elderly relatives, as it helps them respond quickly when assistance may be needed.

Think of it as an additional layer of safety, helping elderly and visually challenged individuals maintain their independence without relying on wearable devices or constant human supervision, while keeping privacy at the centre of its design.