ISAVM: Improved Smart Avian Monitoring using FLANN-based Audio Activity Detection & Speech Enhancement

The Avian monitoring system is one of the challenging tasks that helps identify the environmental changes in the forest as well as the overall counts of specific species. Out of several methods available for avian monitoring, audio-based avian monitoring is one of the most efficient and cost-effective tools. By studying bird sounds, a smart society can be built for an enhanced avian surveillance system through the use of speech recognition algorithms. Conventionally, speech characteristics are employed for these tasks, which may not be appropriate given that these acoustic noises deviate from human speech and deteriorate the identification systems’ performance. An application-specific audio activity identification technique is needed since the features of human voice and bird sound differ. As of now, few works have been reported mainly for the bird sound analysis with audio activity detection and speech enhancement schemes. This work has considered and implemented this problem in three steps. In the first stage, an improved voice activity algorithm is designed using a Functional Link Artificial Neural Network model. In the second stage, an effective AdaBoost classifier is used for training and testing. Finally, the developed model, ISAVM: Improved Smart Avian Monitoring System has been checked for improved performance in two standard bird datasets. The evaluation has been done for different preprocessing options with and without audio activity detection and speech enhancement schemes. It has been observed that the proposed model is performing consistently with more than 93% of classification accuracy which is better than the standard avian monitoring models.