A framework for bioacoustical species classification in a versatile service-oriented wireless mesh network

The decline in amphibian populations worldwide has become a tangible example of a major environmental concern due to the fact that amphibian presence has been interpreted as a good indicator of the health of an ecosystem. Consequently, monitoring is an imperative. In this work, a conceptual framework for bioacoustical species classification is formulated and an instantiation of this framework is presented in the form of a sensor array processing (SAP) system. From a digital signal processing perspective, the main feature implemented in the instantiated SAP is an application capability, based on Mel-frequency cepstrum coefficients (MFCC), principal components analysis (PCA), and k-nearest neighbors (k-NN) methods that allows identifying species from audio vocalizations recorded by array sensors. Finally, the processed information is being delivered, through what has been termed a master sensor node (MSN) configuration, to a versatile service-oriented (VESO) wireless mesh network (WMN) which is currently being implemented as an instrumentation testbed at the National Jobos Bay Estuarine Research Reserve (JBNERR) located on the island of Puerto Rico.