Automatic Activity Monitoring Research Articles

A Scalogram-Based CNN Approach for Audio Classification in Construction Sites

The automatic monitoring of activities in construction sites through the proper use of acoustic signals is a recent field of research that is currently in continuous evolution. In particular, the use of techniques based on Convolutional Neural Networks (CNNs) working on the spectrogram of the signal or its mel-scale variants was demonstrated to be quite successful. Nevertheless, the spectrogram has some limitations, which are due to the intrinsic trade-off between temporal and spectral resolutions. In order to overcome these limitations, in this paper, we propose employing the scalogramas a proper time–frequency representation of the audio signal. The scalogram is defined as the square modulus of the Continuous Wavelet Transform (CWT) and is known as a powerful tool for analyzing real-world signals. Experimental results, obtained on real-world sounds recorded in construction sites, have demonstrated the effectiveness of the proposed approach, which is able to clearly outperform most state-of-the-art solutions.

Relationship between the timing of insemination based on estrus detected by the automatic activity monitoring system and conception rates using sex-sorted semen in Holstein dairy cattle.

We investigated the optimal timing of artificial insemination (AI) for achieving pregnancy according to the onset/end of estrus detected by an accelerometer system in Holstein cattle. The conception rates of conventional semen were used as a reference. The conception rate from AI of sex-sorted semen was higher at −4 to 4 h (57.1%) from the end of estrus than those at −12 to −4 h (37.7%) and 12–20 h (30.3%), whereas AI at 4–12 h showed an intermediate conception rate (47.4%). Conversely, conception rates were similar in AI performed between 0 and 32 h from the onset of estrus. Regarding conventional semen, the interval from the onset and end of estrus did not affect conception rates. The present results suggest that the time of the end of estrus is the better indicator of optimal AI timing for sex-sorted semen than the onset of estrus.

Administration of GnRH at Onset of Estrus, Determined by Automatic Activity Monitoring, to Improve Dairy Cow Fertility during the Summer and Autumn.

Simple SummaryWe used an automatic activity-monitoring system to determine onset of estrus in dairy cows. Within 5 h of onset, we administered a single injection of GnRH analogue to improve fertility during the summer and autumn. The treatment increased pregnancy per insemination during the autumn, but not in the summer. The subgroups for which the treatment specifically tended to improve conception risk during the autumn were: mature (2nd plus parity) cows and cows with uterine disease and ketosis after calving. Detection of estrus onset by activity monitoring and GnRH administration shortly thereafter could be incorporated into a synchronization program, to improve fertility of positively-responding subpopulations of cows.We examined gonadotropin-releasing hormone (GnRH) administration at onset of estrus (OE), determined by automatic activity monitoring (AAM), to improve fertility of dairy cows during the summer and autumn. The study was performed on two dairy farms in Israel. The OE was determined by AAM recorded every 2 h, and a single im dose of GnRH analogue was administered shortly after OE. Pregnancy was determined by transrectal palpation, 40 to 45 d after artificial insemination (AI). Conception risk was analyzed by the GLIMMIX procedure of SAS. Brief visual observation of behavioral estrus indicated that about three-quarters of the events (n = 40) of visually detected OE occurred within 6 h of AAM-detected OE. Accordingly, the GnRH analogue was administered within 5 h of AAM-detected OE, to overlap with the expected endogenous preovulatory LH surge. Overall, pregnancy per AI (P/AI) was monitored over the entire experimental period (summer and autumn) in 233 first, second or third AI (116 and 117 AI for treated and control groups, respectively). Least square means of P/AI for treated (45.8%) and control (39.4%) groups did not differ, but group-by-season interaction tended to differ (p = 0.07), indicating no effect of treatment in the summer and a marked effect of GnRH treatment (n = 58 AI) compared to controls (n = 59 AI) on P/AI in the autumn (56.6% vs. 28.5%, p < 0.03). During the autumn, GnRH-treated mature cows (second or more lactations), and postpartum cows exhibiting metabolic and uterine diseases, tended to have much larger P/AI than their control counterparts (p = 0.07–0.08). No effect of treatment was recorded in the autumn in first parity cows or in uninfected, healthy cows. In conclusion, administration of GnRH within 5 h of AAM-determined OE improved conception risk in cows during the autumn, particularly in those exhibiting uterine or metabolic diseases postpartum and in mature cows. Incorporation of the proposed GnRH treatment shortly after AAM-detected OE into a synchronization program is suggested, to improve fertility of positively responding subpopulations of cows.

Evaluating the association between early-lactation lying behavior and hoof lesion development in lactating Jersey cows

Objectives were to evaluate the association between behavior and hoof lesions (HL) in lactating dairy cows. Jersey cows without any visible HL on all 4 legs were enrolled at 20 ± 3 d in milk (DIM), examined for body condition score (BCS), and had an automatic activity monitor (AfiTag II, AfiMilk, Afikim, Israel) attached to their right hind leg. At 120 ± 3 DIM, activity monitors were removed and cows were re-examined for HL and BCS. Cows were classified according to HL status as either healthy or with HL at d 120. Because sole hemorrhage (SH) accounted for over 80% of HL on d 120, SH was reclassified according to the number of feet affected. Daily activity data (daily lying duration, lying bouts, lying bout duration, and number of steps) of 344 cows collected between d 20 and d 120 were analyzed using restricted maximum likelihood linear mixed models with an autoregressive covariance structure. Separate models were built to include specific activities, HL status at d 120, DIM, interaction between lesion status at d 120 and DIM, parity, season of calving, and BCS change from d 20 to d 120 as fixed effects. Cow was included in all models as random effect. Incidence of HL at d 120 was 58.4% (n = 201 out of 344). Compared with healthy cows, cows with HL at d 120 had reduced daily lying duration (-0.53 h; 95% CI: -0.78 to -0.28 h) in the early postpartum period. We found no evidence for a difference in number of lying bouts (0.41; 95% CI: -0.76 to 1.59;), lying bout duration (-3.04 min; 95% CI: -6.6 to 0.49), and number of steps (62.14; 95% CI: -89.62 to 213.91) between cows that remained healthy and those that developed HL. Compared with healthy cows, cows that developed SH in 1 or more feet had reduced daily lying duration in the early postpartum period. Irrespective of lesion status at d 120, daily lying duration and lying bout duration increased from d 20 to d 120, whereas lying bout number and number of steps decreased from d 20 to d 120. We concluded that cows that developed HL had a reduced daily lying duration in the early postpartum period. Therefore, reduced daily lying duration should be considered a risk factor for HL development in lactating dairy cows.

Evaluation of Feature Extraction and Recognition for Activity Monitoring and Fall Detection Based on Wearable sEMG Sensors.

As an essential subfield of context awareness, activity awareness, especially daily activity monitoring and fall detection, plays a significant role for elderly or frail people who need assistance in their daily activities. This study investigates the feature extraction and pattern recognition of surface electromyography (sEMG), with the purpose of determining the best features and classifiers of sEMG for daily living activities monitoring and fall detection. This is done by a serial of experiments. In the experiments, four channels of sEMG signal from wireless, wearable sensors located on lower limbs are recorded from three subjects while they perform seven activities of daily living (ADL). A simulated trip fall scenario is also considered with a custom-made device attached to the ankle. With this experimental setting, 15 feature extraction methods of sEMG, including time, frequency, time/frequency domain and entropy, are analyzed based on class separability and calculation complexity, and five classification methods, each with 15 features, are estimated with respect to the accuracy rate of recognition and calculation complexity for activity monitoring and fall detection. It is shown that a high accuracy rate of recognition and a minimal calculation time for daily activity monitoring and fall detection can be achieved in the current experimental setting. Specifically, the Wilson Amplitude (WAMP) feature performs the best, and the classifier Gaussian Kernel Support Vector Machine (GK-SVM) with Permutation Entropy (PE) or WAMP results in the highest accuracy for activity monitoring with recognition rates of 97.35% and 96.43%. For fall detection, the classifier Fuzzy Min-Max Neural Network (FMMNN) has the best sensitivity and specificity at the cost of the longest calculation time, while the classifier Gaussian Kernel Fisher Linear Discriminant Analysis (GK-FDA) with the feature WAMP guarantees a high sensitivity (98.70%) and specificity (98.59%) with a short calculation time (65.586 ms), making it a possible choice for pre-impact fall detection. The thorough quantitative comparison of the features and classifiers in this study supports the feasibility of a wireless, wearable sEMG sensor system for automatic activity monitoring and fall detection.

A hierarchical model for recognizing alarming states in a batteryless sensor alarm intervention for preventing falls in older people

Falls are common among older people, especially in hospitals and nursing homes. The combination of pervasive sensing and statistical learning methods is creating new possibilities for automatic monitoring of activities of hospitalized older people to provide targeted and timely supervision by clinical staff to reduce falls. In this paper we introduce a hierarchical conditional random fields model to predict alarming states (being out of the bed or chair) from a passive wearable embodiment of a sensor worn over garment to provide an intervention mechanism to reduce falls. Our approach predicts alarm states in real time and avoids the use of empirically determined heuristics methods alone or in combination with machine learning based models, or multiple cascaded classifiers for generating alarms from activity prediction streams. Instead, the proposed hierarchical approach predicts alarms based on learned relationships between alarms, sensor information and predicted low-level activities. We evaluate the performance of the approach with 14 healthy older people and 26 hospitalized older patients and demonstrate similar or better performance than machine learning based approaches combined with heuristics based methods.

Automated activity monitoring and visual observation of estrus in a herd of loose housed Hereford cattle: Diagnostic accuracy and time to ovulation

A prospective cohort study was performed in the purebred Hereford herd at Götala Beef and Lamb Research Centre, Sweden. The study's first objective was to assess the ability of an automatic activity monitoring system (AAMS) to detect estrus in beef suckler cows, and its second objective was to estimate the time from estrus to ovulation. The study sample (n = 38) consisted of 14 Hereford heifers and 24 Hereford cows. Standardized visual observation of estrus was performed for 20 minutes thrice daily, and animal activity was recorded with an AAMS system, Heatime (SCR Engineers Ltd., Israel). Cows in estrus underwent transrectal ultrasonography every 8 hours, to estimate the time of ovulation. Blood samples for progesterone analysis were collected thrice weekly throughout the study period. A cutoff value of 1-ng progesterone/mL of serum was used to define luteal activity. The AAMS had a 90% (95% confidence interval [CI] 77%–97%) sensitivity and 100% specificity (95% CI 94%–100%), and visual detection of estrus had a 77% sensitivity (95% CI 62%–88%) and a 89% specificity (95% CI 79%–95%) for identifying estrus when compared to the gold standard defined by temporal pattern of serum progesterone concentration. When both methods were used in parallel, the sensitivity increased to 96% (95% CI 86%–99%), and the specificity increased to 90% (95% CI 80%–96%). The time of ovulation after estrus was determined on 50 occasions. The median estrus (AAMS detected) to ovulation interval was 25 hours for heifers and 23 hours for cows (interquartile range 11–29 hours and 19–25 hours, respectively). The median estrus (visually detected) to ovulation interval was 28 hours for heifers and 21 hours for cows (interquartile range 13–29 hours for both categories). In conclusion, the AAMS had both a higher sensitivity and specificity for estrus detection than thrice-daily visual observation. The time from detection of estrus to ovulation observed in this study indicates that reproductive performance might be improved if Hereford cattle are inseminated sooner after detection of estrus than is currently recommended.

The relationship between activity clusters detected by an automatic activity monitor and endocrine changes during the periestrous period in lactating dairy cows

The aim of this study was to determine the relationship between observed estrous-related behavior, activity clusters (AC; detected by automatic activity monitor), endocrine profiles, and ovulation time. Twenty-one cows in estrus (after 2 cloprostenol treatments, 11d apart) and 12 nonsynchronized cows, to establish Heatime (SCR Engineers Ltd., Netanya, Israel) herd baseline activity, were enrolled. Cows had Heatime monitors applied 3 wk before the trial to establish their own baseline activity level. Cows in standing estrus had ultrasonography and phlebotomy carried out every 4h to determine dominant follicle size, endocrine profiles, and ovulation time. After ovulation, these procedures were repeated once on d 3 to 6. Heatime alerted estrus in 90% of cows, and incorrectly alerted 17% of AC. The mean ± SEM duration for standing estrus was 9±1 and 13±1h for estrous-related behavior. Estrous-related behavior began after the start of the proestrous estradiol-17β (E2) increase (59±6.5h). Cows with longer durations of raised proestrous E2 had longer intervals from its onset to the start of standing estrus and AC. The AC duration increased with longer durations of estrous-related behavior. Higher peak E2 occurred with longer standing estrus and estrous-related behavior. As E2 concentration decreased after the peak, 90% of cows still had estrous-related behavior. Duration of estrous-related behavior increased with higher average E2 concentration during the last 8h before the start of the LH surge. During this surge 90% of cows had all of their standing estrus. As yields increased, so did the magnitude of the preovulatory FSH surges. Higher surges occurred with shorter standing estrus and estrous-related behavior. Cows with shorter LH surges had longer standing estrus. Peak LH preceded the AC peak (6.6±0.8h). Duration of overlap between the AC start and the LH surge end ranged between 0 and 14h; 1 cow had none. No association was found between the AC characteristics with the E2, LH, or FSH profiles. In conclusion, the relationship between the timing of the E2 increase and estrous activity may be mediated by other factors (GnRH surge). Estrous-related behavior, but not endocrine profiles, was related to AC duration. Timing of standing estrus during the LH surge ensures that mating allows sperm maturation before ovulation. Based on the interval from the start of an AC to ovulation (27±1h), the optimum time to artificial insemination is, on average, between 9 and 15h after the AC start.

Automatic monitoring of activities of daily living using contactless sensors (AMACS)

Automatic monitoring of activities of daily living using contactless sensors (AMACS)

Segmenting sensor data for activity monitoring in smart environments

Within a smart environment, sensors have the ability to perceive changes of the environment itself and can therefore be used to infer high-level information such as activity behaviours. Sensor events collected over a period of time may contain several activities. The fundamental process of any automatic activity monitoring system is therefore to process streams of sensor events and detect occurrences of activities. In this study, we propose three segmentation algorithms to separate time series sensor data into segments to be further processed by an activity recognition algorithm. A preliminary evaluation of the approaches developed has been conducted on a data set collected from a single person living in an apartment over a period of 28 days. The results show that the proposed approaches can segment sensor data to detect activities and infer sensor segments to recognise activities with accuracies of 81.6, 81.6 and 82.9%, respectively.

Monitoring systems for the support of home care

In this article, the design of a system for the ambient, unobtrusive and automatic monitoring of Activities of Daily Living (ADL) is described. In the context of the growing imbalance between (potentially young) caregivers and (most often older) people receiving care, technical monitoring systems may help to organise care more efficiently and to identify degrading abilities very early to trigger preventive measures. To improve the acceptance of the system described in this article, the selection process of the sensors to be integrated into the flat or to be worn by the older people has been steered by the results of focus group interviews with older people, their relatives and professional caregivers. The interviews revealed that these people would in general accept such systems, but security, mobility and communication aspects have to be clearly and appropriately addressed. In an experimental study the recognition rate of the activity ‘preparation and intake of food or beverages’ has been measured with two age groups (6 subjects, age between 25 and 40/mean 30 years and 5 subjects, age between 72 and 84/mean 75.3 years). The food preparation was detected with a sensitivity of 74.7% and a specificity of 84.2% using a vision sensor.

Behavioral deficits in adult rats treated neonatally with glutamate

The present study evaluated long-term behavioral consequences of neonatal monosodium- l-glutamate (MSG) treatment in rats. The pups received MSG (3 mg/g sc) daily from postnatal day (PD) 5–12. Data from an automatic activity monitor showed that locomotion of MSG-treated females and males aged 56 and 84 days was significantly reduced. Beginning PD 120, three behavioral tests were performed. As compared to the controls, in the elevated plus maze test, modified to evaluate the adaptive form of spatial memory, MSG-treated animals of both sex had significantly prolonged start and transfer latencies. In the social recognition test, assessing olfactory working memory, MSG-treated males displayed a reduced interest in the juvenile conspecific as the stimulus partner during both the initial exposure and re-exposure performed 30 min later. In the open field test, a significant decrease in the habituation rate was found in MSG-treated animals. Sex-dependent differences in behavioral performance were suggested in the open field and elevated plus maze tests. Behavioral changes are discussed in light of the deficits in perception and processing of visual and olfactory stimuli.

Measuring daily behavior using ambulatory accelerometry: the Activity Monitor.

Advanced ambulatory systems that measure aspects of overt human behavior during normal daily life have become feasible, owing to developments in data recording and sensor technology. One such instrument is the Activity Monitor (AM). This paper provides a technical description of the AM and information about its validity and current applications. The AM is based on ambulatory accelerometry, the aim of which is to assess postures and motions for long-term (> 24-h) measurement periods during normal daily life. Accelerometers are attached to the thighs, trunk, and lower arms, and signals are continuously stored in a digital portable recorder. In the postmeasurement analysis, postures and motions are detected by means of custom-made software programs. Validity studies performed on different populations showed high agreement scores between the computerized and automatic AM output and the visually analyzed video recordings. The AM has so far been applied in rehabilitation, psychophysiology, and cardiology but has many possibilities in behavioral research.

Nocturnal and diurnal activity patterns and roosting sites of green sandpipers Tringa ochropus wintering in southern England

A radio telemetric study of Green Sandpipers at their wintering site, a disused watercress bed in southern England, shows that they alternate between the watercress bed and a nearby gravel pit complex over the course of the autumn and winter. Birds spent most days feeding at the watercress bed and roosted overnight in the gravel pit complex. During extremely cold weather in January and February the birds switched to roosting at the watercress beds. Automatic activity monitoring showed that the birds were active for around 80% of each day at all times of year. Their time active at night varied from around 16% in autumn to over 40% in cold conditions in midwinter. The hours the birds were active during the day and night in mid‐winter were inversely proportional to the maximum daily temperature. The evidence suggests that a low level of night time activity is normal in Green Sandpipers but the high levels found during extremely cold winter conditions are the result of birds attempting to increase their daily food intake thus supporting the ‘supplementary’ hypothesis for nocturnal foraging.

An inexpensive automatic activity monitor for small primates

An activity cage has been designed to provide a quantitative measure of locomotor activity in small primates. Movement is detected by the interruptions of photocell beams, which are counted electronically and stored in on-board memory for subsequent display or entry into a computer. The device functions independently, and does not require a computer to be on-line during data collection. Data thus obtained have proven reliable, and are a valid record of the activity of the animal.