A method for automatic detection of sleep apnea using pulse photopletismography signal (PPG) is proposed. This method is based on a detection of decreases on PPG amplitude fluctuations. The proposed detector is composed of three stages: pre-processing, envelope detection, based on root mean square series or Hilbert transform, and decision algorithm based on an adaptive threshold. The detector has been evaluated using simulated and real signals. Sensibility and positive predictive value of the detector where 76% and 73% for real signals. A clinical study to sleep apnea diagnosis in children based on this detector has been carried out. PPG attenuation events per hour ratio Eh has statistical significance (p < 0.05) to classify children as normal 13.5 ± 6.35 E h (mean ± SD) or pathologic 21.1 ± 8.93 E h .