Birralee International School Trondheim

More than 2.2 billion people worldwide have impaired near or distance vision, with about 4.95 million people, including 240,000 children, in India being blind and 70 million visually impaired people. These individuals find it hard to navigate through an unfamiliar environments and to identify objects. Existing navigation aids are usually bulky and can’t be used on a day-to-day basis. Smart Glasses were developed to improve the lives of visually impaired people by enabling them to navigate their surroundings effectively. The device comprises a USB camera mounted on a pair of eyeglasses for live video capturing of the user’s surroundings, a Raspberry Pi 4 running a Python program utilizing YOLOv10 for detecting objects while the MiDaS model estimates the distance of the detected objects, and audio feedback about that through an earphone to the user. MiDaS performs monocular depth estimation using its encoder-decoder architecture, ResNet-based backbone, use of skip connections, and multi-scale feature fusion. Smart Glasses was tested in a sequence of controlled tests to determine its capabilities in object detection, of various commonly encountered objects. It managed to achieve an overall detection accuracy of about 93%. For specific objects such as apple, person, and car, specifically, detection accuracy was 94%, whereas for tables, benches and cellphones, it was 95%. The detection accuracy for a chair was 90%. Detection accuracy for objects in the category of transport (motorcycles, buses, trucks and stop signs) ranged between 92 and 94%, ensuring outdoor navigation safety. The depth estimation was also found to be reliable, with proximity of the nearest objects estimated correctly 90% of the time, and for the farthest, it was 94%, confirming the system’s aptitude for spatial awareness. These results indicate Smart Glasses can empower the navigation capability of visually impaired individuals and hence exceeded traditional assistive devices. Five completely blind people took part in the testing of the device at Devnar School for the Blind, Hyderabad. Each participant wore the glasses for 10 minutes and was asked to describe the experience of understanding their surroundings during that period. The users appreciated the glasses for their daily use and requested features such as book reading, multilingual audio feedback (Telugu and Hindi), and facial recognition.

With the increase of urbanization around the world, key habitats such as grasslands and scrublands are disappearing, posing significant threats to species that rely on such habitats. Passive acoustic monitoring (PAM) has started to gain interest as a reliable tool for biodiversity monitoring. This study assesses whether PAM can be an effective way to monitor biodiversity with relatively low effort when compared to traditional survey methods. Using PAM, a mixed grassland and scrubland ecosystem was monitored for five months for avifaunal diversity. Data that was collected from dusk to dawn using audio recorders deployed on the study site were run through BirdNET analyzer under default detectors. Top detections for each species were compared with avian vocalization libraries manually. Out of 135 species detected, 76 were true positives, resulting in a species identification accuracy of 56.3%. Notably, among the species that were true detections, 21 species were found to have confidence scores larger than 0.99. In addition, 15 of the detected species were found to be migratory in this area, and three were found to be rare. This study demonstrates how PAM can be used to monitor biodiversity in a species-rich but understudied area such as Southern India, to identify both cryptic and nocturnal species that might be omitted from standard field surveys. ### Competing Interest Statement The authors have declared no competing interest. TIFR Hyderabad, RTI4007 DBT Ramalingaswami Re-entry Fellowship, BT/RLF/Re-entry/07/2022

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder affecting more than 75 million people in the world. Early detection is crucial for reducing lifetime care costs and improving the quality of life of the individual with Autism. Current diagnostic methods rely heavily on behavioral observations and thus lack the precision required for early detection. The study explores the use of speech combined with machine learning for early Autism detection. The study utilizes ASDBank and CHILDES, clinical databases within TalkBank, to obtain audio recordings of children with and without ASD. The children were instructed to complete the same task to ensure consistency and fairness across the trials. Auditory features- Spectral Centroid, Spectral Rolloff, Zero Crossing Rate, and Root Mean Square-are first extracted from the audio files. Then we compared 3 models-Support Vector Machine (SVM), Random Forest (RF) classifier, and a Neural Network (NN)- to determine the most appropriate model for autism detection using the 4 auditory features. The SVM model achieved the highest accuracy. The second part of the paper focuses on an ablation study to determine which of the features is the most impactful for this detection. The results indicated that Zero Crossing Rate was the most impactful feature.

Japanese third-age English language learners, though growing, are still an under-researched demographic. Grounded on the quantitative results of Japanese third-age learners' English language proficiency (ELP), which is B1 on average, this descriptive-qualitative study explored why the participants got certain levels of ELP. Using the maximum variation sampling, the researchers chose five third-aged Japanese participants from the five cities of Aichi, Japan. The findings reveal two significant themes: motivational factors (with three sub-themes: personal interest, social engagement, and personal enrichment) and factors affecting ELP (with four sub-themes: dynamics of sex and ELP, dynamics of age and ELP, dynamics of education and ELP, dynamics of time and ELP), encourage the examination of learning a language as a means to greater autonomy, cultural sensitivity, and an understanding of the global community. It provides valuable insights for teachers or curriculum designers to improve EFL programs for third-age learners, enhancing accessibility, interest, and effective language acquisition.

Autism spectrum disorder (ASD) is a neurodevelopmental condition affecting one in thirty-six children. It often emerges early but is diagnosed later, making early detection crucial for not only improving quality of life but also reducing lifetime care costs. The current diagnostic methods in India, relying primarily on behavioural observations, lack the precision and early detection capabilities needed. This paper explores the use of non-intrusive neuroimaging techniques, specifically structural magnetic resonance imaging (sMRI) and functional magnetic resonance imaging (fMRI) combined with machine learning (ML) for early ASD detection in India. Western nations usually conduct these neuroimaging investigations, identifying structural and functional brain abnormalities such as increased brain volume and altered white matter integrity in people with ASD. However, traditional neuroimaging analysis is labour intensive and prone to human error. In contrast, medical practitioners in India rely more on their tacit knowledge and intuitive intelligence for ASD diagnosis rather than neuroimaging. This research makes use of a Partial Least Squares Structural Equation Modelling (PLS-SEM) to analyse feedback from medical practitioners to a questionnaire focused on the integration of ML and neuroimaging for ASD diagnosis in India. It uses a 5-point Likert scale to answer the research questions leading to the conclusion, common challenges, as well as recommendations on the future steps. We conclude that ML-based diagnostic tools, through its adaptation and application for the Indian context, will assist medical practitioners in diagnosing ASD in India.