With the aggravating trend of aging of population, the population aged over 60 years of age is growing faster than any other age groups. Under these circumstances, the number of elderly people living alone is increasing. Therefore, there is increasing expectation for elderly monitoring services which can detect their general activities or emergency situations such as having a fall. Keeping the daily fundamental activities of the elderly is also important to prevent future accidents. In this paper, from the perspective of general versatility and privacy, an activity recognition method using a low-resolution infrared array sensor is proposed. The system can overcome the limitation of exist methods such as an invasion of privacy and have a wider application. A long short-term memory classifier is applied to this system in order to improve the accuracy. Experiments show that the system successfully achieves the aim of higher-precision human motion detection.