Design and development of keyword identification system
The "keyword identification" project is defined with the aim of designing a target word identification system with the ability to quickly and appropriately adapt to the change of the unknown target word. In this report, the activities of the project's technical team and their results are discussed. In order to train and test the system introduced in the first step, the multilingual database of spoken words has been reviewed by the technical team of the project and is ready to be used precisely. In addition to this database, various data augmentation methods have been used to enrich it. In the following, by examining the most up-to-date methods and structures introduced in this field, the final and proposed models are presented, which are trained and evaluated using this database and the few-shot learning method. In this report, this model has been examined in terms of accuracy and speed on continuous sound and its results have been presented in detail.