This paper aims to study the feasibility of using asmart mobile phone with an embedded accelerometer in gaitpattern monitoring. The second motivation is to examine theimpact of the accelerometer sampling frequency on gait analysis.A mobile phone and a standalone accelerometer sensorwere simultaneously attached to subject’s lower back to recordwalking patterns. The degree of agreement between gait featuresderived from two devices was assessed in terms ofaverage error rate, normalised limits of agreement and intraclasscorrelation. Various agreement levels were observed forthree temporal features, three root mean square features, fiveregularity features and two symmetry features. The downsamplingdata were used to examine the impact of sampleintervals on the gait features. Eleven out of 13 features havenormalised mean difference less than 0.1 when sample intervalswere less than 50ms. To carry out a further evaluation, thefeatures derived from the downsampling gait data were usedto classify subjects with chronic pain and health subjects, anda classification accuracy of 90% was achieved. The resultsshowed that it is feasible and reliable to assess and monitorgait patterns based on spatio-temporal gait features derivedfrom smart mobile phones with an embedded accelerometer.