Invited review: The evolution of cattle bioacoustics and application for advanced dairy systems

Calming Hungarian Grey Cattle in Headlocks Using Processed Nasal Vocalization of a Mother Cow

Sound analysis is an important field of research for improving precision livestock farming systems. If the information carried by livestock sounds is interpreted correctly, it could be used to improve management and welfare assessment in this field. Therefore, we hypothesized that the nasal vocalization of a mother cow could have a calming effect on conspecifics. The nasal vocalization in our study was recorded from a mother cow (not part of the test herd) while it was licking its day-old calf. The raw sound was analyzed, cleaned from noises, and the most representative vocalization was lengthened to two minutes. Thirty cows having calves were randomly selected from eighty Hungarian grey cattle cows. Two test days were selected, one week apart; the weather circumstances in both days were similar. The herd was collected in a paddock, and the test site (a restraining crate with a headlock) was 21 m away from them. The cows from the herd were gently moved to the restraining crate, and, after the installation of the headlock, Polar® heart rate monitors were fixed on the animals. The recording of the RR intervals was carried out for two minutes. On day one of the test, the processed nasal sound was played to every second cow during the heart rate monitoring. When the sound ended, the heart rate monitor was removed. On test day two, the sound and no sound treatments were switched among the participating cows. At the end of the measurement, the headlock was opened, letting the animals out voluntarily, and a flight test was performed along a 5 m distance. The time needed to pass the 5 m length was measured with a stopwatch and divided by the distance. The RR intervals were analyzed with the Kubios HRV Standard (ver. 3.5.0) software. The following data were recorded for the entire measurement: average and maximum heart rate; SD1 and SD2; pNN50; VLF, LF, and HF. The quasi-periodic signal detected in the sound analyses can hardly be heard, even when it is enhanced to the maximum. This can be considered a vibration probably caused by the basis of articulation, such as a vibration of the tongue, for example. The SD2/SD1 ratio (0.97 vs. 1.07 for the animals having no sound and sound played, respectively, p = 0.0110) and the flight speed (0.92 vs. 1.08 s/m for the animals having no sound and sound played, respectively, p = 0.0409) indicate that the sound treatment had a calming effect on the restrained cows. The day of the test did not influence any of the measured parameters; therefore, no effect of the routine was observed. The yes-no sequence of the sound treatment significantly reduced the pNN50 and flight speed values, suggesting a somewhat more positive association with the headlock and the effectiveness of the processed nasal sound. In conclusion, we have demonstrated that, by means of sound analyses, not only information about individuals and the herd can be gathered but that, with proper processing, the sound obtained can be used to improve animal welfare.

Automated identification of chicken distress vocalizations using deep learning models

The annual global production of chickens exceeds 25 billion birds, which are often housed in very large groups, numbering thousands. Distress calling triggered by various sources of stress has been suggested as an 'iceberg indicator' of chicken welfare. However, to date, the identification of distress calls largely relies on manual annotation, which is very labour-intensive and time-consuming. Thus, a novel convolutional neural network-based model, light-VGG11, was developed to automatically identify chicken distress calls using recordings (3363 distress calls and 1973 natural barn sounds) collected on an intensive farm. The light-VGG11 was modified from VGG11 with significantly fewer parameters (9.3 million versus 128 million) and 55.88% faster detection speed while displaying comparable performance, i.e. precision (94.58%), recall (94.89%), F1-score (94.73%) and accuracy (95.07%), therefore more useful for model deployment in practice. To additionally improve light-VGG11's performance, we investigated the impacts of different data augmentation techniques (i.e. time masking, frequency masking, mixed spectrograms of the same class and Gaussian noise) and found that they could improve distress calls detection by up to 1.52%. Our distress call detection demonstration on continuous audio recordings, shows the potential for developing technologies to monitor the output of this call type in large, commercial chicken flocks.

Systematic review of animal-based indicators to measure thermal, social, and immune-related stress in pigs

The intense nature of pig production has increased the animals’ exposure to stressful conditions, which may be detrimental to their welfare and productivity. Some of the most common sources of stress in pigs are extreme thermal conditions (thermal stress), density and mixing during housing (social stress), or exposure to pathogens and other microorganisms that may challenge their immune system (immune-related stress). The stress response can be monitored based on the animals’ coping mechanisms, as a result of specific environmental, social, and health conditions. These animal-based indicators may support decision making to maintain animal welfare and productivity. The present study aimed to systematically review animal-based indicators of social, thermal, and immune-related stresses in farmed pigs, and the methods used to monitor them. Peer-reviewed scientific literature related to pig production was collected using three online search engines: ScienceDirect, Scopus, and PubMed. The manuscripts selected were grouped based on the indicators measured during the study. According to our results, body temperature measured with a rectal thermometer was the most commonly utilized method for the evaluation of thermal stress in pigs (87.62%), as described in 144 studies. Of the 197 studies that evaluated social stress, aggressive behavior was the most frequently-used indicator (81.81%). Of the 535 publications examined regarding immune-related stress, cytokine concentration in blood samples was the most widely used indicator (80.1%). Information about the methods used to measure animal-based indicators is discussed in terms of validity, reliability, and feasibility. Additionally, the introduction and wide spreading of alternative, less invasive methods with which to measure animal-based indicators, such as cortisol in saliva, skin temperature and respiratory rate via infrared thermography, and various animal welfare threats via vocalization analysis are highlighted. The information reviewed was used to discuss the feasible and most reliable methods with which to monitor the impact of relevant stressors commonly presented by intense production systems on the welfare of farmed pigs.

Convolutional Neural Networks for the Identification of African Lions from Individual Vocalizations

The classification of vocal individuality for passive acoustic monitoring (PAM) and census of animals is becoming an increasingly popular area of research. Nearly all studies in this field of inquiry have relied on classic audio representations and classifiers, such as Support Vector Machines (SVMs) trained on spectrograms or Mel-Frequency Cepstral Coefficients (MFCCs). In contrast, most current bioacoustic species classification exploits the power of deep learners and more cutting-edge audio representations. A significant reason for avoiding deep learning in vocal identity classification is the tiny sample size in the collections of labeled individual vocalizations. As is well known, deep learners require large datasets to avoid overfitting. One way to handle small datasets with deep learning methods is to use transfer learning. In this work, we evaluate the performance of three pretrained CNNs (VGG16, ResNet50, and AlexNet) on a small, publicly available lion roar dataset containing approximately 150 samples taken from five male lions. Each of these networks is retrained on eight representations of the samples: MFCCs, spectrogram, and Mel spectrogram, along with several new ones, such as VGGish and stockwell, and those based on the recently proposed LM spectrogram. The performance of these networks, both individually and in ensembles, is analyzed and corroborated using the Equal Error Rate and shown to surpass previous classification attempts on this dataset; the best single network achieved over 95% accuracy and the best ensembles over 98% accuracy. The contributions this study makes to the field of individual vocal classification include demonstrating that it is valuable and possible, with caution, to use transfer learning with single pretrained CNNs on the small datasets available for this problem domain. We also make a contribution to bioacoustics generally by offering a comparison of the performance of many state-of-the-art audio representations, including for the first time the LM spectrogram and stockwell representations. All source code for this study is available on GitHub.

Effects of dietary and milking frequency changes and administration of cabergoline on clinical udder characteristics in dairy cows during dry-off

We investigated the effects of 2 diet energy densities [normal lactation diet (NORM) vs. energy-reduced diet (REDU), both fed for ad libitum intake] and 2 daily milking frequencies [twice (2×) vs. once (1×)] during 1 week before the dry-off day, as well as effects of an injection of either a dopamine agonist [cabergoline (CAB); Velactis, Ceva Santé Animale, Libourne, France; labelled for use only with abrupt dry-off, e.g. no reduction in diet energy density or milking frequency before the last milking] or saline (SAL) following the last milking, on clinical udder characteristics of Holstein cows. During a week before and after the last milking, the following measures were recorded: palpation-based udder firmness and soreness; image-based hock-hock distance; responsiveness to mechanical udder stimulation and degree of udder fill measured with a dynamometer. Before the last milking, REDU cows displayed lower odds of having a firm udder and lower degree of udder fill, as well as lower responsiveness to mechanical udder stimulation, than NORM cows. After the last milking, REDU cows displayed shorter hock-hock distance compared with NORM cows. The effects of milking frequency on the clinical udder characteristics were unclear. Within 24 h following injection, CAB cows showed lower odds of having a firm udder, shorter hock-hock distance, and lower degree of udder fill than SAL cows, irrespective of treatment group before dry-off. In this study, reducing diet energy density prior to dry-off, and to some extent administering the dopamine agonist cabergoline after the last milking, resulted in fewest clinical udder changes.

Review: Anticipating alternative trajectories for responsible Agriculture 4.0 innovation in livestock systems

Technological change has been a constant feature of livestock systems leading to the third agricultural 'green' revolution of the mid-20th century. Digital technologies are now leading us into the fourth agricultural revolution, where sustainable food production is supported by technologies that collect data useful for farm and supply chain performance improvement, along with task automation and compliance. However, the potential benefits of digital agricultural futures are uncertain and plagued by unrealized expectations of previous innovations. The aims of this paper are to articulate current trends in technology and livestock systems and anticipate future trajectories for Agriculture 4.0 in relation to meeting sustainability and animal welfare outcomes for livestock systems. We use a 'Futures Triangle' approach to review the role of technology in livestock systems. The main findings are that previous work envisioning technological livestock futures have favoured pull of the future factors (techno-optimists) or weight of the past (techno-pessimists), rather than a balance of pull, push and weighting factors. Responsible Agriculture 4.0 innovation requires public-private collaboration of innovation system stakeholders, including policy makers, farmers, consumers, as well as technology developers, to enable development of transition pathways from a systems perspective. The use of responsible innovation processes, including anticipation on alternative futures, should also be built into innovation processes to support critical reflection on technological trajectories and related innovation system consequences, both desirable and undesirable.

Assessing effects of dietary and milking frequency changes and injection of cabergoline during dry-off on hunger in dairy cows using 2 feed-thwarting tests

We investigated the single and combined effects of 2 feeding levels (normal lactation diet vs. energy-reduced diet, both fed for ad libitum intake) and 2 daily milking frequencies (twice vs. once) during 1 wk before the dry-off day (d 0), as well as an intramuscular injection of either a dopamine agonist (cabergoline; Velactis, Ceva Santé Animale; labeled for use only with abrupt dry-off, e.g., no reduction in feeding level or milking frequency before the last milking) or saline after the last milking on d 0 on the feeding motivation of clinically healthy, loose-housed, pregnant, lactating Holstein cows. From d 0, all cows were fed the same dry-cow diet for ad libitum intake. Cows were subjected to 2 feed-thwarting tests, a test in the home pen using their diets (test A: d -6, -1, and 1; during 35 min when the feed bins were filled, but locked) and another test carried out in an adjacent pen in which access to concentrate provided in a familiar plastic box was blocked by a wire-mesh lid (test B: d -5 and 2). In test A, we recorded how often cows attempted to feed per 35 min, whether cows vocalized during the 35-min period, and latency to feed within 300 s after feed bins were unlocked. In test B, we recorded latency to approach either of 2 familiar boxes (the wire-mesh box and an identical open box with a small portion of concentrate) within 600 s and how often cows directed behaviors toward the wire-mesh box (number of occurences/5 min). On d -6 (test A), no clear differences in feeding motivation among treatments were found. On d -5 and -1, cows fed the energy-reduced diet displayed a higher probability of vocalizing (test A), were more than 50% quicker to feed (test A), were approximately 5× quicker to approach a box (test B), and directed 60% more behavior toward the wire-mesh box (test B) than cows fed the normal diet. Moreover, cows fed the energy-reduced diet attempted to feed approximately 75% more on d -1 compared with d -6 (test A). On d 2 (test B), cows previously fed the normal diet directed 40% more behavior toward the wire-mesh box than cows previously fed the energy-reduced diet. Reducing feeding level, either before or on the dry-off day, resulted in consistently increased feeding motivation, interpreted as a sign of hunger. No clear effects of change in milking frequency, singly or combined with reduced diet energy density, on feeding motivation were found before d 0. Whereas, on d 2, cows previously milked twice daily were quicker to approach a box than cows previously milked once daily. Cows injected with cabergoline attempted to feed more, but showed lower probability of vocalizing compared with saline-injected cows (d 1; test A), irrespective of treatment before d 0. The effects of cabergoline on feeding motivation are not easily interpreted and warrant further investigation. From a hunger perspective, reducing milking frequency rather than diet energy density seems to be a less negative management to reduce milk production before dry-off.

The effects of heat stress on the behaviour of dairy cows – a review

Heat stress in livestock is a function of macro- and microclimatic factors, their duration and intensity, the environments where they occur and the biological characteristics of the animal. Due to intense metabolic processes, high-producing dairy cows are highly vulnerable to the effects of heat stress. Disturbances in their thermoregulatory capability are reflected by behavioural, physiological and production changes. Expression of thermoregulatory behaviour such as reduction of activity and feed intake, searching for a cooler places or disturbances in reproductive behaviours may be very important indicators of animal welfare. Especially maintain of standing or lying position in dairy cattle may be a valuable marker of the negative environmental impact. Highly mechanized farms with large numbers of animals have the informatic system can detect alterations automatically, while small family farms cannot afford these type of equipments. Therefore, observing and analysing behavioural changes to achieve a greater understanding of heat stress issue may be a key factor for developing the effective strategies to minimize the effects of heat stress in cattle. The aim of this review is to present the state of knowledge, over the last years, regarding behavioural changes in dairy cows (Bos Taurus) exposed to heat stress conditions and discuss some herd management strategies provided mitigation of the overheat consequences.

The effect of rearing conditions during the milk-fed period on milk yield, growth, and maze behaviour of dairy cows during their first lactation

The objective was to find whether cow growth, milk performance, and behaviour are affected by (1) rearing conditions until weaning after a milk-fed period of 84 d and (2) the sire lineage. Thirty-five Holstein heifers were assigned to one of three treatments: SM, n=13, pen with mother to 21st day, then group pen (they received a maximum of 6 kg of milk daily); SN, n=9, after 3 d with own mother in pen with nursing cow (they received a maximum of 6 kg of milk daily); H, n=13, in hutch from the 2nd to 56th day (6 kg of milk replacer daily), then loose housing pen to weaning (6 kg of milk replacer daily). After weaning at the 84th day, all heifers were kept in pens with the same ration as during calving. During lactation, live body weight (LBW) was measured each month and milk yield each day. Maze learning was evaluated in the fifth month of lactation. The data were analysed using a general linear model ANOVA. At the 30th day, the LBW tended to be the highest in SN (SM 528.2 ± 11.4 kg, SN 571.7 ± 15.3 kg, H 533.2 ± 12.3 kg). When lactation ended, the highest LBW was in SN and the lowest in H (SM 612.6 ± 12.2 kg, SN 623.1 ± 16.4 kg, H 569.8 ± 13.2 kg; P<0.05). The SN tended to have the highest production of milk (SM 7143.9 ± 241.5 kg, SN 7345.1 ± 319.0 kg, H 7146.7 ± 234 kg), and the H for FCM (SM 6290.3 ± 203.2 kg, SN 6307.6 ± 268.4 kg, H 6399.3 ± 197.1 kg) for 305 d lactation. Group SN crossed the maze fastest (SM 1141.4 ± 120.5 s, SN 810.3 ± 160.5 s, H 1120.8 ± 118.6 s). The vocalization number differed significantly (SM 32.3 ± 5.7, SN 20.8 ± 4.4, H 9.9 ± 2.6; P<0.01). The results indicated that the rearing method up to weaning may have an impact on dairy cows' performance and behaviour.

Vocal production in postpartum dairy cows: Temporal organization and association with maternal and stress behaviors

Mammalian vocalizations can encode contextual information in both the spectrographic components of their individual vocal units and in their temporal organization. Here we observed 23 Holstein-Friesian dairy cows immediately after birth during interactions with their calf and when their calf was separated to the other side of a fence line. We investigated whether the vocalizations emitted in these postpartum contexts would vary temporally. We also described the maternal and stress behaviors preceding and following postpartum vocal production using kinematic diagrams and characterized call sequence structure. The kinematic diagrams highlight the disruption of maternal responses caused by calf separation and show that behavioral and vocal patterns varied according to the cows' emotional states and proximity to the calf in both contexts. During calf interactions, cows mainly produced closed-mouth calls simultaneous to licking their calf, whereas an escalation of stress responses was observed during calf separation, with the cows approaching the fence line, becoming alert to the calf, and emitting more mixed and open-mouth calls. Call sequences were similarly structured across contexts, mostly containing repetitions of a single call type, with a mean interval of 0.57 s between calls and a greater cumulative vocalization duration, attributed to an increased number of vocal units per sequence. Overall, calf separation was associated with a greater proportion of calls emitted as a sequence (inverse of single isolated calls), a shorter interval between separate call sequences, and a greater number of vocal units per sequence, compared with calf interactions. These temporal vocal features varied predictably with the high stress expression from cows during calf separation and may represent temporal modulations of emotional expression. Despite the noisy farm soundscape, empirical call type and temporal vocal features were easy to measure; thus, findings could be applied to future cattle studies wishing to analyze vocalizations for on-farm welfare assessments.

Vocal individuality of Holstein-Friesian cattle is maintained across putatively positive and negative farming contexts

Cattle mother-offspring contact calls encode individual-identity information; however, it is unknown whether cattle are able to maintain individuality when vocalising to familiar conspecifics over other positively and negatively valenced farming contexts. Accordingly, we recorded 333 high-frequency vocalisations from 13 Holstein-Friesian heifers during oestrus and anticipation of feed (putatively positive), as well as denied feed access and upon both physical and physical & visual isolation from conspecifics (putatively negative). We measured 21 source-related and nonlinear vocal parameters and stepwise discriminant function analyses (DFA) were performed. Calls were divided into positive (n = 170) and negative valence (n = 163) with each valence acting as a 'training set' to classify calls in the oppositely valenced 'test set'. Furthermore, MANOVAs were conducted to determine which vocal parameters were implicated in individual distinctiveness. Within the putatively positive 'training set', the cross-validated DFA correctly classified 68.2% of the putatively positive calls and 52.1% of the putatively negative calls to the correct individual, respectively. Within the putatively negative 'training set', the cross-validated DFA correctly assigned 60.1% of putatively negative calls and 49.4% of putatively positive calls to the correct individual, respectively. All DFAs exceeded chance expectations indicating that vocal individuality of high-frequency calls is maintained across putatively positive and negative valence, with all vocal parameters except subharmonics responsible for this individual distinctiveness. This study shows that cattle vocal individuality of high-frequency calls is stable across different emotionally loaded farming contexts. Individual distinctiveness is likely to attract social support from conspecifics, and knowledge of these individuality cues could assist farmers in detecting individual cattle for welfare or production purposes.

Symposium review: Scientific assessment of affective states in dairy cattle

Affective states, which refer to feelings or emotions, are a key component of animal welfare, but these are also difficult to assess. Drawing upon a body of theoretical and applied work, we critically review the scientific literature on the assessment of affective states in animals, drawing examples where possible from research on dairy cattle, and highlighting the strengths and weaknesses of scientific methods used to assess affective states in animals. We adopt the "valence/arousal" framework, describing affect as a 2-dimensional space (with valence referring to whether an experience is positive or negative, and arousal referring to the intensity of the experience). We conclude that spontaneous physiological and behavioral responses typically reflect arousal, whereas learned responses can be valuable when investigating valence. We also conclude that the assessment of affective states can be furthered using mood assessments and that the use of drug treatments with known emotional effects in humans can be helpful in the assessment of specific affective states in animals.

An Appraisal of Methods for Measuring Welfare of Grazing Ruminants

Although disturbances in body function of animals can be measured to determine whether a state of stress may exist, there is growing interest in finding ways to assess their emotional status as an indicator of good or bad welfare status. Generally it is easier to determine poor states of well-being than positive ones. For grazing ruminants some indicators of well-being include absence of illness, good growth and productivity, and longevity. Motion detectors can provide automated remote monitoring of behavior and it is likely that there will be advances in the interpretation software to increase the utility of this technology for assessing well-being. Cortisol levels in body fluids, feces and pelage are prominent as a marker of poor animal welfare, but like many of the other objective measures that are used, are not wholly reliable at the individual animal level. These other measures include: plasma serotonin, heart rate variation, infra-red thermography, cytokines, salivary alpha amylase, and acute phase proteins. Use of automated facial expression recognition may supplement electrophysiological recording as means to quantify the pain experience of animals. Although the measures described in the literature do not necessarily provide the final answer for determination of welfare in grazing ruminants, they all have some merit and deserve further investigation.

Automatic recording of individual oestrus vocalisation in group-housed dairy cattle: development of a cattle call monitor

Oestrus detection remains a problem in the dairy cattle industry. Therefore, automatic detection systems have been developed to detect specific behavioural changes at oestrus. Vocal behaviour has not been considered in such automatic oestrus detection systems in cattle, though the vocalisation rate is known to increase during oestrus. The main challenge in using vocalisation to detect oestrus is correctly identifying the calling individual when animals are moving freely in large groups, as oestrus needs to be detected at an individual level. Therefore, we aimed to automate vocalisation recording and caller identification in group-housed dairy cows. This paper first presents the details of such a system and then presents the results of a pilot study validating its functionality, in which the automatic detection of calls from individual heifers was compared to video-based assessment of these calls by a trained human observer, a technique that has, until now, been considered the 'gold standard'. We developed a collar-based cattle call monitor (CCM) with structure-borne and airborne sound microphones and a recording unit and developed a postprocessing algorithm to identify the caller by matching the information from both microphones. Five group-housed heifers, each in the perioestrus or oestrus period, were equipped with a CCM prototype for 5 days. The recorded audio data were subsequently analysed and compared with audiovisual recordings. Overall, 1404 vocalisations from the focus heifers and 721 vocalisations from group mates were obtained. Vocalisations during collar changes or malfunctions of the CCM were omitted from the evaluation. The results showed that the CCM had a sensitivity of 87% and a specificity of 94%. The negative and positive predictive values were 80% and 96%, respectively. These results show that the detection of individual vocalisations and the correct identification of callers are possible, even in freely moving group-housed cattle. The results are promising for the future use of vocalisation in automatic oestrus detection systems.

Automated bioacoustics: methods in ecology and conservation and their potential for animal welfare monitoring

Vocalizations carry emotional, physiological and individual information. This suggests that they may serve as potentially useful indicators for inferring animal welfare. At the same time, automated methods for analysing and classifying sound have developed rapidly, particularly in the fields of ecology, conservation and sound scene classification. These methods are already used to automatically classify animal vocalizations, for example, in identifying animal species and estimating numbers of individuals. Despite this potential, they have not yet found widespread application in animal welfare monitoring. In this review, we first discuss current trends in sound analysis for ecology, conservation and sound classification. Following this, we detail the vocalizations produced by three of the most important farm livestock species: chickens ( Gallus gallus domesticus), pigs ( Sus scrofa domesticus) and cattle ( Bos taurus). Finally, we describe how these methods can be applied to monitor animal welfare with new potential for developing automated methods for large-scale farming.