Campus violence is a common social phenomenon all over the world, and is the most harmful type of school bullying events. As artificial intelligence and remote sensing techniques develop, there are several possible methods to detect campus violence, e.g., movement sensor-based methods and video sequence-based methods. Sensors and surveillance cameras are used to detect campus violence. In this paper, the authors use image features and acoustic features for campus violence detection. Campus violence data are gathered by role-playing, and 4096-dimension feature vectors are extracted from every 16 frames of video images. The C3D (Convolutional 3D) neural network is used for feature extraction and classification, and an average recognition accuracy of 92.00% is achieved. Mel-frequency cepstral coefficients (MFCCs) are extracted as acoustic features, and three speech emotion databases are involved. The C3D neural network is used for classification, and the average recognition accuracies are 88.33%, 95.00%, and 91.67%, respectively. To solve the problem of evidence conflict, the authors propose an improved Dempster–Shafer (D–S) algorithm. Compared with existing D–S theory, the improved algorithm increases the recognition accuracy by 10.79%, and the recognition accuracy can ultimately reach 97.00%.

School bullying is a serious problem among teenagers. School violence is one type of school bullying and considered to be the most harmful. As AI (Artificial Intelligence) techniques develop, there are now new methods to detect school violence. This paper proposes a video-based school violence detecting algorithm. This algorithm first detects foreground moving targets via the KNN (K-Nearest Neighbor) method and then preprocesses the detected targets via morphological processing methods. Then, this paper proposes a circumscribed rectangular frame integrating method to optimize the circumscribed rectangular frame of moving targets. Rectangular frame features and optical-flow features were extracted to describe the differences between school violence and daily-life activities. We used the Relief-F and Wrapper algorithms to reduce the feature dimension. SVM (Support Vector Machine) was applied as the classifier, and 5-fold cross validation was performed. The accuracy was 89.6%, and the precision was 94.4%. To further improve the recognition performance, we developed a DT–SVM (Decision Tree–SVM) two-layer classifier. We used boxplots to determine some features of the DT layer that are able to distinguish between typical physical violence and daily-life activities and between typical daily-life activities and physical violence. For the remainder of activities, the SVM layer performed a classification. For this DT–SVM classifier, the accuracy reached 97.6%, and the precision reached 97.2%, thus showing a significant improvement.

Functional near-infrared spectroscopy (fNIRS) is commonly used as a non-invasive tool to measure cerebral neurovascular dynamics. Its potential for diagnostics of various brain disorders has been already demonstrated in many recent studies, including Alzheimer’s disease (AD). fNIRS studies are usually based on comparing hemoglobin measurements at baseline and during a specific task. At present, many proposed methods using fNIRS to diagnose AD involve certain tasks, which may be challenging for the elderly and patients with cognitive decline. Here, we propose a method to characterize AD patients and control in resting state, by applying spectral entropy (SE) analysis on oxyhemoglobin and deoxyhemoglobin, HbO and HbR, respectively, and total hemoglobin (HbT) based on fNIRS signals measured from the left and right sides of the forehead. We applied SE to very low frequency (VLF) (0.008–0.1 Hz), respiratory (0.1–0.6 Hz), and cardiac (0.6–5 Hz) bands to find out which band delivered the optimum result. Next, a t test with 0.05 significant level was performed to compare SE values of AD patients and controls. Results from the VLF band looked promising as SE values from AD patients were always significantly higher than those from controls. In addition, this phenomenon was consistent for both sides of the forehead. However, significant differences in SE values in the respiratory band were found from the left hemisphere only, and in the cardiac band from the right hemisphere only. SE value from the VLF band supports a strong argument that it provides good predictability related to the development of AD. We demonstrated that SE of brain fNIRS signal can be an useful biomarker for Alzheimer’s disease pathology.

Functional near-infrared spectroscopy (fNIRS) has been utilized already around three decades for monitoring the brain, in particular, oxygenation changes in the cerebral cortex. In addition, other optical techniques are currently developed for in vivo imaging and in the near future can be potentially used more in human brain research. This paper reviews the most common label-free optical technologies exploited in brain monitoring and their current and potential clinical applications. Label-free tissue monitoring techniques do not require the addition of dyes or molecular contrast agents. The following optical techniques are considered: fNIRS, diffuse correlations spectroscopy (DCS), photoacoustic imaging (PAI) and optical coherence tomography (OCT). Furthermore, wearable optical brain monitoring with the most common applications is discussed.

In-vivo microscopical studies indicate that brain cerebrospinal fluid (CSF) transport driven by blood vessel pulsations is reduced in the early stages of Alzheimer’s disease (AD). We hypothesized that the coupling pattern between cerebrovascular pulsations and CSF is altered in AD, and this can be measured using multi-wavelength functional near-infrared spectroscopy (fNIRS). To study this, we quantified simultaneously cerebral hemo- and CSF hydrodynamics in early AD patients and age-matched healthy controls. Physiological pulsations were analysed in the vasomotor very low frequency (VLF 0.008–0.1 Hz), respiratory (Resp. 0.1–0.6 Hz), and cardiac (Card. 0.6–5 Hz) bands. A sliding time window cross-correlation approach was used to estimate the temporal stability of the cerebrovascular-CSF coupling. We investigated how the lag time series variation of the coupling differs between AD patients and control. The couplings involving deoxyhemoglobin (HbR) and CSF water, along with their first derivative, in the cardiac band demonstrated significant difference between AD patients and controls. Furthermore, the lag time series variation of HbR-CSF in the cardiac band provided a significant relationship, p-value = 0.04 and r² = 0.16, with the mini-mental state exam (MMSE) score. In conclusion, the coupling pattern between hemodynamics and CSF is reduced in AD and it correlates with MMSE score.

Measuring cardiac activity from the chest using an accelerometer is commonly referred to as seismocardiography. Unfortunately, it cannot provide clinically valid data because it is easily corrupted by motion artefacts. This paper proposes two methods to improve peak detection from noisy seismocardiography data. They rely on discrete wavelet transform analysis using either biorthogonal 3.9 or reverse biorthogonal 3.9. The first method involves slicing chest vibrations for each cardiac activity, and then detecting the peak location, whereas the other method aims at detecting the peak directly from chest vibrations without segmentation. Performance evaluations were conducted on signals recorded from small children and adults based on missing and additional peaks. Both algorithms showed a low error rate (15.4% and 2.1% for children/infants and adults, respectively) for signals obtained in resting state. The average error for sitting and breathing tasks (adults only) was 14.4%. In summary, the first algorithm proved more promising for further exploration.