Background Methods for faller classification and gait patterns using portable and reliable equipment are still a challenge when considered an active and healthy aging population, in particular for large-scale applications. In this study, we investigated the patterns that could identify elderly faller by means of the acceleration gait data. Method Gait frequency pattern study of active older than 60 years participants divided into three groups according to the fall history in the last year: 10 non-fallers, 10 sporadic fallers and 10 recurrent fallers. The subjects performed 6MWT, with a triaxial accelerometer allocated at the waist, with 200 Hz sampling rate. Magnitude information was extracted to analyze the acceleration curves. The comparison between the groups was performed through the chi-squared test, ANOVA and Kruskal-Wallis test. Results Average age of the participants was 75.13±6.37 years (76.6% women). While it was not possible to differentiate fallers through the 6MWT, statistical differences were found in variables extracted from segmented acceleration curves. The first and third largest frequency amplitudes measured when participants are turning during the test showed discriminative capability between sporadic fallers and the other groups (p=0.018; 0.014, respectively). Statistical difference was found also from the difference between the maximum magnitude in frequency of the complete signal in relation to the turning movement between recurrent fallers and the remaining groups (p=0.035), while the third main frequency magnitude difference between walking and turning showed significant difference between sporadic fallers and the others (p=0.030). The best result considering ROC analysis achieved AUC of 0.807, sensitivity of 0.80 and specificity of 0.75. Conclusion and Relevance Frequency features extracted from the combination of the accelerometer and the test can contribute to clinical practice and to scientific area, providing a low-cost diagnosis with a wearable sensor, allowing large-scale applications. Considering specific gait events are important for this task since our results indicate there is more discriminative capability when comparing the turning movement with other gait patterns.