A methodological approach to compare continuous and instantaneous sampling and two methods to deal with animals out of sight on dairy cattle behavior and interaction with their calf in the first hours post-partum.

Animal behavior can provide useful information about animal welfare, but methods and tools used to gather behavioral data and data treatment can influence the results. Therefore, this study was carried out on dairy cow (Bos taurus) behavior and interaction with calves early post-partum aiming at comparing two sampling rules, namely continuous and instantaneous sampling at scan intervals of 30 s, 1, 2, 3, 4, 5, and 10 min, and of two methods to deal with out of sight animals. The study was based on three assumptions: (1) continuous sampling provides the most complete and accurate data, allowing the observation of seldom behaviors and short events; (2) instantaneous sampling can provide accurate measurements of frequency and duration, especially at short scan intervals; (3) differences in behavioral results may occur depending on whether a correction for time out of sight is applied or not. Thus, 10 dams were observed from videos in the 2 h post-parturition. Ruminating, stereotypies, calf-biting and calf-butting were not recorded during the observation period. Other behaviors were observed only with continuous sampling or with continuous and instantaneous at 30-s scan intervals. The recoding of several behaviors was less accurate when applying longer scan intervals. Data from continuous and instantaneous sampling at 30-s scan intervals were compared with Wilcoxon test. Results showed no significant differences for posture, position in the pen and all behaviors (p > 0.05) except vocalizing (p = 0.003). The same test did not highlight significant differences due to method of dealing with out of sight for both sampling rules (p > 0.05). Correlation between continuous and instantaneous sampling were prevalently high at 30-s intervals and they decreased as the length of scan intervals increased for most behaviors. Results confirmed the first two assumptions suggesting that continuous sampling is more accurate, in particular for short and rare behaviors, and caution against the suitability of dam behavioral data collected using instantaneous sampling at scan intervals of minutes. The third assumption was not proven by this study. Results should be considered in light of the development of new technologies that relies on data acquired by sensors and imaging to monitor cow-calf welfare and behavior post-parturition.

Explainable Siamese Neural Network for Classifying Pediatric Respiratory Sounds.

The field of medical acoustics is gaining constantly-increasing attention by the scientific community, with the the general goal being the automatic understanding of medical related signals to assist medical personnel in the decision making process. In this direction, this work introduces a framework able to differentiate between normal and abnormal respiratory sounds by learning relationships characterizing pairs of sounds. More specifically, considering the nature of respiratory sounds, we designed a feature set able to capture the coarse and fine structure exhibited such signals by means of multiresolution analysis. Similar/dissimilar relationships are modeled via a suitably-learned Siamese Neural Network encompassing a series of convolutional layers. Interestingly, such a relationship learning framework conveniently solves the existing class imbalance problem as it is trained on pairs of similar/dissimilar audio signals. Importantly, we employed the dataset designed for the IEEE BioCAS 2022 Grand challenge on Respiratory Sound Classification along with a standardized experimental protocol allowing reproducibility and reliable comparison between different approaches. After extensive experiments assessing the proposed framework from diverse points of view, including an ablation study, it is shown that it outperforms existing approaches, while providing explainable predictions via a Q&A scheme allowing interaction with the medical experts.

A perceptual measure for evaluating the resynthesis of automatic music transcriptions.

This study focuses on the perception of music performances when contextual factors, such as room acoustics and instrument, change. We propose to distinguish the concept of "performance" from the one of "interpretation", which expresses the "artistic intention". Towards assessing this distinction, we carried out an experimental evaluation where 91 subjects were invited to listen to various audio recordings created by resynthesizing MIDI data obtained through Automatic Music Transcription (AMT) systems and a sensorized acoustic piano. During the resynthesis, we simulated different contexts and asked listeners to evaluate how much the interpretation changes when the context changes. Results show that: (1) MIDI format alone is not able to completely grasp the artistic intention of a music performance; (2) usual objective evaluation measures based on MIDI data present low correlations with the average subjective evaluation. To bridge this gap, we propose a novel measure which is meaningfully correlated with the outcome of the tests. In addition, we investigate multimodal machine learning by providing a new score-informed AMT method and propose an approximation algorithm for the p-dispersion problem.

Edge Computing for Vision-Based, Urban-Insects Traps in the Context of Smart Cities.

Our aim is to promote the widespread use of electronic insect traps that report captured pests to a human-controlled agency. This work reports on edge-computing as applied to camera-based insect traps. We present a low-cost device with high power autonomy and an adequate picture quality that reports an internal image of the trap to a server and counts the insects it contains based on quantized and embedded deep-learning models. The paper compares different aspects of performance of three different edge devices, namely ESP32, Raspberry Pi Model 4 (RPi), and Google Coral, running a deep learning framework (TensorFlow Lite). All edge devices were able to process images and report accuracy in counting exceeding 95%, but at different rates and power consumption. Our findings suggest that ESP32 appears to be the best choice in the context of this application according to our policy for low-cost devices.

What's in a Meow? A Study on Human Classification and Interpretation of Domestic Cat Vocalizations.

Although the domestic cat (Felis catus) is probably the most widespread companion animal in the world and interacts in a complex and multifaceted way with humans, the human-cat relationship and reciprocal communication have received far less attention compared, for example, to the human-dog relationship. Only a limited number of studies have considered what people understand of cats' human-directed vocal signals during daily cat-owner interactions. The aim of the current study was to investigate to what extent adult humans recognize cat vocalizations, namely meows, emitted in three different contexts: waiting for food, isolation, and brushing. A second aim was to evaluate whether the level of human empathy toward animals and cats and the participant's gender would positively influence the recognition of cat vocalizations. Finally, some insights on which acoustic features are relevant for the main investigation are provided as a serendipitous result. Two hundred twenty-five adult participants were asked to complete an online questionnaire designed to assess their knowledge of cats and to evaluate their empathy toward animals (Animal Empathy Scale). In addition, participants had to listen to six cat meows recorded in three different contexts and specify the context in which they were emitted and their emotional valence. Less than half of the participants were able to associate cats' vocalizations with the correct context in which they were emitted; the best recognized meow was that emitted while waiting for food. Female participants and cat owners showed a higher ability to correctly classify the vocalizations emitted by cats during brushing and isolation. A high level of empathy toward cats was significantly associated with a better recognition of meows emitted during isolation. Regarding the emotional valence of meows, it emerged that cat vocalizations emitted during isolation are perceived by people as the most negative, whereas those emitted during brushing are perceived as most positive. Overall, it emerged that, although meowing is mainly a human-directed vocalization and in principle represents a useful tool for cats to communicate emotional states to their owners, humans are not particularly able to extract precise information from cats' vocalizations and show a limited capacity of discrimination based mainly on their experience with cats and influenced by empathy toward them.

Automatic Classification of Cat VocalizationsEmitted in Different Contexts.

Cats employ vocalizations for communicating information, thus their sounds can carry a widerange of meanings. Concerning vocalization, an aspect of increasing relevance directly connected withthe welfare of such animals is its emotional interpretation and the recognition of the production context.To this end, this work presents a proof of concept facilitating the automatic analysis of cat vocalizationsbased on signal processing and pattern recognition techniques, aimed at demonstrating if the emissioncontext can be identified by meowing vocalizations, even if recorded in sub-optimal conditions. Werely on a dataset including vocalizations of Maine Coon and European Shorthair breeds emitted in threedifferent contexts: waiting for food, isolation in unfamiliar environment, and brushing. Towards capturing theemission context, we extract two sets of acoustic parameters, i.e., mel-frequency cepstral coefficients andtemporal modulation features. Subsequently, these are modeled using a classification scheme based ona directed acyclic graph dividing the problem space. The experiments we conducted demonstrate thesuperiority of such a scheme over a series of generative and discriminative classification solutions. Theseresults open up new perspectives for deepening our knowledge of acoustic communication betweenhumans and cats and, in general, between humans and animals.

Automatic acoustic classification of insect species based on directed acyclic graphs.

This work presents the design of a directed acyclic graph (DAG) scheme, the nodes of which incorporate hidden Markov models (HMMs) for classifying insect species. Such a DAG scheme is able to limit the problem space, while having the HMMs capture the temporal evolution of Mel-scaled spectrograms extracted out of wingbeat sounds. Interestingly, the proposed approach offers interpretability of the classification process by inspecting the sequence of edges activated in the DAG (path). The dataset encompasses 50 000 wingbeat sounds representing six species, i.e., Ae. aegypti (male and female), Cx. quinquefasciatus (male and female), Cx. stigmatosoma (male and female), Cx. tarsalis (male and female), Musca domestica, and Drosophila simulans, and is publicly available at https://sites.google.com/site/insectclassification/. Thorough species classification experiments showed that the proposed solution outperforms state-of-the-art approaches.

Transfer Learning for Improved Audio-Based Human Activity Recognition.

Human activities are accompanied by characteristic sound events, the processing of which might provide valuable information for automated human activity recognition. This paper presents a novel approach addressing the case where one or more human activities are associated with limited audio data, resulting in a potentially highly imbalanced dataset. Data augmentation is based on transfer learning; more specifically, the proposed method: (a) identifies the classes which are statistically close to the ones associated with limited data; (b) learns a multiple input, multiple output transformation; and (c) transforms the data of the closest classes so that it can be used for modeling the ones associated with limited data. Furthermore, the proposed framework includes a feature set extracted out of signal representations of diverse domains, i.e., temporal, spectral, and wavelet. Extensive experiments demonstrate the relevance of the proposed data augmentation approach under a variety of generative recognition schemes.

A transfer learning framework for predicting the emotional content of generalized sound events.

Predicting the emotions evoked by generalized sound events is a relatively recent research domain which still needs attention. In this work a framework aiming to reveal potential similarities existing during the perception of emotions evoked by sound events and songs is presented. To this end the following are proposed: (a) the usage of temporal modulation features, (b) a transfer learning module based on an echo state network, and (c) a k-medoids clustering algorithm predicting valence and arousal measurements associated with generalized sound events. The effectiveness of the proposed solution is demonstrated after a thoroughly designed experimental phase employing both sound and music data. The results demonstrate the importance of transfer learning in the specific field and encourage further research on approaches which manage the problem in a synergistic way.

Fault Identification in Distributed Sensor Networks Based on Universal Probabilistic Modeling.

This paper proposes a holistic modeling scheme for fault identification in distributed sensor networks. The proposed scheme is based on modeling the relationship between two datastreams by means of a hidden Markov model (HMM) trained on the parameters of linear time-invariant dynamic systems, which estimate the specific relationship over consecutive time windows. Every system state, including the nominal one, is represented by an HMM and the novel data are categorized according to the model producing the highest likelihood. The system is able to understand whether the novel data belong to the fault dictionary, are fault-free, or represent a new fault type. We extensively evaluated the discrimination capabilities of the proposed approach and contrasted it with a multilayer perceptron using data coming from the Barcelona water distribution network. Nine system states are present in the dataset and the recognition rates are provided in the confusion matrix form.

A cognitive fault diagnosis system for distributed sensor networks.

This paper introduces a novel cognitive fault diagnosis system (FDS) for distributed sensor networks that takes advantage of spatial and temporal relationships among sensors. The proposed FDS relies on a suitable functional graph representation of the network and a two-layer hierarchical architecture designed to promptly detect and isolate faults. The lower processing layer exploits a novel change detection test (CDT) based on hidden Markov models (HMMs) configured to detect variations in the relationships between couples of sensors. HMMs work in the parameter space of linear time-invariant dynamic systems, approximating, over time, the relationship between two sensors; changes in the approximating model are detected by inspecting the HMM likelihood. Information provided by the CDT layer is then passed to the cognitive one, which, by exploiting the graph representation of the network, aggregates information to discriminate among faults, changes in the environment, and false positives induced by the model bias of the HMMs.