Elephant presence detection using NN based audio pattern recognition
In recent years, the continued existence of wild elephants have come under severe threat due to habitat loss along with the human-elephant conflict. The available solutions are mostly based on active mechanisms such as wearing tracking collars on the elephants and deploying electric fences. Although these solutions work up to some context, they are not sustainable solutions. Therefore, the requirement for a passive solution rise-up. The proposed solution is an elephant detection system using audio patterns
The system is self-powered as it harnesses energy using a solar panel. Therefore, the low power operation is essential. A single detection node uses 3 PIR sensors to detect motion within the vicinity. Once a motion is detected, it triggers the MAX78000's CNN accelerator to detect the patterns in the audio feed when there is a presence of elephants. Rather than always keeping the CNN on and keeping the MAX78000 in high power stages reduce the life span of the batteries significantly. Therefore, this mechanism allows the system to run on battery for longer timeseven though the MAX78000 can perform this AI task using very low power when compared to other options available in the market to run a neural network at the edge.
Once the presence is detected after processing the audio feed, the node indicates the presence of the elephants using a long range Sub-1GHz link to a distant receiver. The sub -1GHz runs on low data rates to extend range and to operate on low power. The distant receiver takes necessary actions to evade the elephants or reduce the speed of a train or a vehicle as a precaution to prevent a collision with an elephant. As a single detection node cannot achieve the expected results, there will be several nodes deployed in a vicinity of monitoring. Each detection node can communicate over a BLE Mesh network among each other. The single detection node consists of a Ti's multiprotocol wireless CC1352 MCU along with MAX78000.
Once the presence is detected after processing the audio feed, the node indicates the presence of the elephants using a long range Sub-1GHz link to a distant receiver. The sub -1GHz runs on low data rates to extend range and to operate on low power. The distant receiver takes necessary actions to evade the elephants or reduce the speed of a train or a vehicle as a precaution to prevent a collision with an elephant. As a single detection node cannot achieve the expected results, there will be several nodes deployed in a vicinity of monitoring. Each detection node can communicate over a BLE Mesh network among each other. The single detection node consists of a Ti's multiprotocol wireless CC1352 MCU along with MAX78000.
Discussion (0 comments)