Kenichi Tanaka

Learning motion patterns and anomaly detection by Human trajectory analysis

In this paper, we propose a novel method to learn motion patterns and detect anomalies by human trajectory analysis. Human trajectories are various, for example, moving, roaming, pausing, and so on. But, current approaches for the analysis of motion patterns are effective only in understanding simple trajectories. We aim to understand complicated human trajectories with long-term observation. To deal with spatial and temporal features of trajectories, we employ HMM (Hidden Markov Model) to model time-series features of human positions. Next, a similarity matrix of HMM mutual distances is formed. MDS (Multi-Dimensional Scaling) based on eigenvector decomposition provides projected coordinates of trajectories in low-dimensional space. Then we apply k-means clustering to projected data in order to acquire human motion patterns. Anomalies can be detected by the use of likelihood scores for HMM representing motion patterns. We tested the proposed method by real-world trajectories data observed in a small store. Experimental result shows that our method accurately finds typical motion patterns and unusual trajectories.

Wide-area Surveillance Sensor with Leaky Coaxial Cables

This paper presents a microwave sensing system for wide-area surveillance that comprises two leaky coaxial cables, one for transmitting and the other for receiving. The sensor detects an intrusion by measuring the electromagnetic field level, which fluctuates when an intruder approaches and passes through the surveillance area. In this research, we have succeeded in specifying an intrusion point by applying the spread-spectrum technique. Results from a field test prove that the detection algorithm could detect the motion of an intruder without experiencing interference from shaking trees or rainfall

Pattern recognition for video surveillance and physical security

We introduce our pattern recognition technologies for video surveillance and physical security. Firstly we explain the system to detect violent actions in an elevator car. The system detects the occurrence of a relevant action via image motion analysis. Secondly we briefly describe our face recognition algorithms by Adaboost. Generally speaking, any face recognition system has operational difficulties caused by facial occlusions, expressions, or aging. Focusing on facial occlusions, we are developing a face recognition system using auto-associative memory. In this paper we show some applied results. In conclusions, we clarify our future directions.

Wide Area Intruder Detection System with a Pair of Transceiver Cables

We have developed a wide area intruder detection surveillance system which was delivered to the Japanese market last summer. This technology is known as MELWATCH which stands for MELco WATCH(to keep lookout). It has been installed at several sites in Japan. MELWATCH is composed of a pair of sensor cables, a sensor unit and an intrusion detection terminal. The sensor cables are composed of a transmission and a receiving cable (transceiver cables), both of which are off-the-shelf leaky coaxial cables. The sensor cables are deployed in parallel along the boundary to be monitored. Upon activation, the sensor unit sends an electric wave into the transmission cable which results in the electric wave leaks from the cable. The receiving cable receives the leaked electric wave and an electric field is formed between the cables. When an intruder enters the electric field, the sensor unit detects this activity and the precise location of the intrusion. Moreover, this sensor is less affected by the environmental noise, such as trees or small animals, nor weather conditions. In addition, the ability to specify the intrusion point makes it possible to configure the more sophisticated system with pan-tilt-zoom cameras. In this paper, we show the effectiveness of MELWATCH through the field data of the several installed sites.

An Apparatus For Measuring Eye And Head Movements Using Ccd Image Sensor And 3space

New apparatus to measure the head and eye movements in real time has been developed. System is used as an instrument to diagnose vertigo patients. Eye movement is observed by a charge coupled device (CCD) image sensor camera. Motion of head is detected by 3SPACE ISOTRAK, of which working principle is measuring orthogonal magnetic fields. Data of three normal and three abnormal subjects including vertiginous persons have been acquired by the system.

Visuospatial working memory game and measured memory performances at various ages

Because of rapid population aging, it is necessary to design and develop senior-friendly or disability-friendly interfaces that can decrease the cognitive workload caused by an interface. At that time, the design implications and evaluation criteria of an interface should be needed for creating senior-friendly and disability-friendly interfaces. One of the elements that relate to memory functions for manipulating interfaces include working memory. However, rough standard of visuospatial memory remains unknown. Particularly, there are little reports about the relations between the age and the volume of visuospatial memory. In this paper, we aim to clarify this relations for proposing interface implications by using a visual pattern span test. For effective measurement of the memory, we implemented a gamified measurement application based on visual pattern span tests. The results indicated that the median numbers of memorable buttons on visuospatial memory were and 7.0 (interquartile range: 5.0–9.0). Also, the number of memorable buttons increases as the age increases until the age ranged 21–25 and then decreases gradually as the age increases after 21–25 years. Our evaluation suggest that it can be effective to measure memory performance of, especially, children by the applications that are designed based on the entertainment concepts including a gamification.

Visuospatial Workload Measurement of an Interface Based on a Dual Task of Visual Working Memory Test

Guidelines devised to ensure that multiple and highly functional in-vehicle information systems (IVISs) are designed to present a distraction-free interface at the same time, need to consider the safety, functionality, and usability of image displays. However, no guidelines exist for quantitatively assessing drivers cognitive workloads, and this includes usability and memorability. In this paper, we propose a method to measure drivers visuospatial workload quantitatively. The method is incorporated in various interface tests including driving and human machine interface evaluations. The results indicated that the success rates of visuospatial working memory tasks can be used to rank the relative difficulties of the tasks, and the ranking order of these rates tended to be similar to that of the subjective difficulties.

Recording Eye Movement During Stepping in Place With a CCD (Charge-Coupled Device) Imagesensor

A new system for recording eye movements using a charge-coupled device (CCD) imagesensor in a video camera, in which less than 0.5 degree of horizontal and vertical eye movements can be recorded, was applied in 7 subjects. Eye movements were recorded while they stepped in place. We observed 5 normal subjects, a patient with horizontal and vertical oscillopsia during walking, and one with a chronic limp. In the normal adults, the amplitude of vertical eye movement was less than 0.5 degree, and the horizontal eye movement was less than 2 degrees. The mean amplitude of horizontal and vertical eye movements in the subject who limped was larger than that of normal subjects. The patient who complained of vertical oscillopsia during walking showed the largest amplitude of vertical eye movement among any subject tested. The vertical oscillopsia in the patient was due to the largest amplitude of vertical eye movement. This system for recording eye movements will be useful in analyzing the mechanism of gaze fixation during walking.

Measuring system of eye movement and posture for vertiginous patients

We have developed a new system using a 3SPACE ISOTRAK, a charge coupled device (CCD) image sensor system and a force plate to measure motion of the head, the eye and the center of gravity of the body in real time.