The present study aimed to apply the Generalized Wavelet Coherence Analysis (GWCA) technique to evaluate whether the variability of knee-ankle joint coupling during gait in elderly fallers and non-fallers is related to the incidence of falls, as compared to young people. Kinematic data were collected from the lower limbs of 30 young people, 17 non-fallers, and 13 elderly fallers while walking at three speeds: the preferred walking speed (PWS), 120% PWS, and 80% PWS. All participants exercise regularly, at least three times a week. They followed a protocol of three 1-min walks on a treadmill, at each speed, for data collection in random order. Kinematic data from 16 reflective markers were used to calculate knee and ankle joint angles in four gait phases (first double stance, single stance, second double stance, and swing), from which coordinative variability was calculated in the sagittal plane. The results showed that the GWCA technique could distinguish the gait of the groups of elderly who fell and did not fall. There were statistical differences for the speed factor, with the greatest variability in 80% and 120% of the PWS, with predominant dominance of the ankle under the knee. Another purpose of the work is to show how much GWCA can be used as an additional tool for gait analysis as it complements the analysis of gait phases, which is already widespread in the literature, and evaluates the intensity of coupling between the joints.