: Practical application of a human monitoring system draws increasing attention with the high developments of information & communication technology and robot technology. Utilizing these technologies, a kindergarten support system for children, parents, and childminders using the activity recognition (AR) has been studied in author's previous studies. In this paper, several wake/sleep identification algorithms for infants as an additional AR were verified by using orthodox algorithms that were used for adults. Investigating combinations of sensor position to infant's body (arm or leg), sampling frequency (5 Hz to 50 Hz), threshold for acceleration detection (0.1 m/s 2 to 0.9 m/s 2 ), and types of wake/sleep detection algorithms (count-scaled, Cole, Sadeh, and zero-threshold algorithms), it was confirmed that the count-scaled algorithm could obtain highest identification accuracy as high as 89.0% under adequate measurement condition.