EVALUATING CLASSROOM ACTIVITIES USING LOW-COST SENSORS AND IOT TECHNOLOGIES

One basic area of concern for many schools is what happens inside a classroom. While many classroom observation instruments like CLASS, ISI and Stallings has been developed to observe a classroom, most of these instruments are expensive to implement because they involve an observer painstakingly recording and coding activities in the classroom. Classroom videos have also been used to record and code teachers' behaviour. However, most of these methods do not scale due to the high cost. This is especially true for developing countries. This paper explores the use of Internet of Things (IoT) to build an automated system for classroom behaviour classification. Voice patterns in a classroom collected using a cheap mic are associated with episodes of interactive or non-interactive classroom behaviour. Arduino and Raspberry Pi microcontrollers are employed as edge devices to manage data that is communicated using the MQTT protocol. An NSQL database called CouchDB is used as the primary storage. Various classifiers including Naive Bayes, Multi-Layer Perceptron, J-Rip and J48 are evaluated for categorizing patterns of audio activity into classroom behaviours. A cross-platform mobile application providing access to all the relevant reports for a variety of stakeholders like teachers, principals and regulatory bodies is also described.