Photoplethysmogram (PPG) signal reflects blood volume changes in peripheral vascular system and can be used to derive multitude of surrogate cardiovascular measurements, including respiration. Under ambulatory monitoring and stress-exercises, PPG signal is prone to corruption by motion artifacts (MA), leading to measurement inaccuracies. In this paper, we propose a method based on combination of variational mode decomposition (VMD) and ensemble empirical mode decomposition (EEMD) to estimate the respiration rate (RR) from motion corrupted PPG. The signal was decomposed using VMD to identify the various frequency components and heart rate, followed by extraction of amplitude, baseline and frequency modulation due to respiration using EEMD. Finally, the accurate estimation of RR was done from these three components. To test and validate, we used Multiparameter Intelligent Monitoring in Intensive Care (MIMIC)-II database and volunteers' data collected at our laboratory. Results of our method showed mean absolute error (MAE) of 0.41 breaths/min for 10 subjects from volunteers' data and 0.35 breaths/min over 53 subjects from MIMIC-II database which is better than the existing methods in terms of accuracy. © 2019 IEEE.