Efficient Speech Detection in Environmental Audio Using Acoustic Recognition and Knowledge Distillation.

The ongoing biodiversity crisis, driven by factors such as land-use change and global warming, emphasizes the need for effective ecological monitoring methods. Acoustic monitoring of biodiversity has emerged as an important monitoring tool. Detecting human voices in soundscape monitoring projects is useful both for analyzing human disturbance and for privacy filtering. Despite significant strides in deep learning in recent years, the deployment of large neural networks on compact devices poses challenges due to memory and latency constraints. Our approach focuses on leveraging knowledge distillation techniques to design efficient, lightweight student models for speech detection in bioacoustics. In particular, we employed the MobileNetV3-Small-Pi model to create compact yet effective student architectures to compare against the larger EcoVAD teacher model, a well-regarded voice detection architecture in eco-acoustic monitoring. The comparative analysis included examining various configurations of the MobileNetV3-Small-Pi-derived student models to identify optimal performance. Additionally, a thorough evaluation of different distillation techniques was conducted to ascertain the most effective method for model selection. Our findings revealed that the distilled models exhibited comparable performance to the EcoVAD teacher model, indicating a promising approach to overcoming computational barriers for real-time ecological monitoring.

Mapping of the Acoustic Environment at an Urban Park in the City Area of Milan, Italy, Using Very Low-Cost Sensors.

The-growing influence of urbanisation on green areas can greatly benefit from passive acoustic monitoring (PAM) across spatiotemporal continua to provide biodiversity estimation and useful information for conservation planning and development decisions. The capability of eco-acoustic indices to capture different sound features has been harnessed to identify areas within the Parco Nord of Milan, Italy, characterised by different degrees of anthropic disturbance and biophonic activity. For this purpose, we used a network of very low-cost sensors distributed over an area of approximately 20 hectares to highlight areas with different acoustic properties. The audio files analysed in this study were recorded at 16 sites on four sessions during the period 25-29 May (2015), from 06:30 a.m. to 10:00 a.m. Seven eco-acoustic indices, namely Acoustic Complexity Index (ACI), Acoustic Diversity Index (ADI), Acoustic Evenness Index (AEI), Bio-Acoustic Index (BI), Acoustic Entropy Index (H), Normalized Difference Soundscape Index (NSDI), and Dynamic Spectral Centroid (DSC) were computed at 1 s integration time and the resulting time series were described by seven statistical descriptors. A dimensionality reduction of the indices carrying similar sound information was obtained by performing principal component analysis (PCA). Over the retained dimensions, describing a large (∼80%) variance of the original variables, a cluster analysis allowed discriminating among sites characterized by different combination of eco-acoustic indices (dimensions). The results show that the obtained groups are well correlated with the results of an aural survey aimed at determining the sound components at the sixteen sites (biophonies, technophonies, and geophonies). This outcome highlights the capability of this analysis of discriminating sites with different environmental sounds, thus allowing to create a map of the acoustic environment over an extended area.

Applications and advances in acoustic monitoring for infectious disease epidemiology.

Emerging infectious diseases continue to pose a significant burden on global public health, and there is a critical need to better understand transmission dynamics arising at the interface of human activity and wildlife habitats. Passive acoustic monitoring (PAM), more typically applied to questions of biodiversity and conservation, provides an opportunity to collect and analyse audio data in relative real time and at low cost. Acoustic methods are increasingly accessible, with the expansion of cloud-based computing, low-cost hardware, and machine learning approaches. Paired with purposeful experimental design, acoustic data can complement existing surveillance methods and provide a novel toolkit to investigate the key biological parameters and ecological interactions that underpin infectious disease epidemiology.