Self-attention-based multimodality convolutional volcano eruption network based indoor localization and wayfinding for blind people

The ability to navigate and precisely locate a visually challenged people within a building is significantly important for a wide variety of location services and public safety. Also, Indoor Localization (IL) for blind people is an open research challenge and the individual localization algorithm failed to achieve accurate location results. To overcome this problem, in this research work, a novel hybrid indoor localization strategy using a WiFi fingerprint database is proposed. It includes two stages such as Offline Phase (OP) and Online Phase (OnP). In the OF, this system employs a Min-Max Normalization approach to normalize the fingerprint database. Then, the Fractional Sparse Fuzzy C-Means Clustering method (FSFCC) is proposed to retrieve the Validation Set (VS) from all the Training Sets (TS) in the ratio of 4:1. In the online stage, the received signal strength indicator (RSSI) data value in dBm is reconstructed and key features such as longitude, latitude, floor ID and building ID are extracted by using Self-Attention-Based Multimodality Convolutional Volcano Eruption Network (Se-AMCVE). Hence, the weight parameters of the proposed network are optimized by applying Volcano Eruption Optimization (VEO). Finally, the accurate estimated location data is the outcome of the proposed network which helps the blind to find the path. The achieved accuracy is 94%, mean positioning error is 3.01 m and floor hit rate is 97.08% which is implemented in the python flask framework.