Real-time monitoring of the movement of elderly people can provide information about an individual's physical movement, degree of functional ability, and general level of activity. This paper describes the real-time monitoring of fall detection in elderly persons using acceleration data. We used a necklace tag device consisting of a Ultra high frequency (UHF) low-power RF system-on-chip and a triaxial accelerometer sensor. The necklace tag processes activity data using the triaxial accelerometer data and classifies types of risk postures, such as stumbling, forward fall, backward fall, and side fall, by calculating signal vector magnitude, angle, angular velocity, power, and spectral energy using a fast Fourier transform (FFT) technique. Because data processing takes place inside the necklace tag, the use of a centralised computer server is no longer necessary. Nevertheless, for daily recordings of activity, zone, and accident time, it is still important to store data at a data centre and to connect them with the patient's medical history. A scheme for continuous daily activity monitoring and fall detection was developed and had a recognition accuracy of more than 95%, demonstrating excellent fall detection and the feasibility of using the proposed method for daily activity monitoring.