With the advent of miniature sensor technology, it is now possible to collect data on various aspects of human movement under free-living conditions. This technology has the potential to be used in activity monitoring systems in several areas, including health, military, sports applications, and human monitoring. The majority of research in wearables technology is focused on skin-mounted sensors or embedded in tight clothes. However, most of our daily clothes are loose or contain wide parts. This paper is interested in analyzing measurements of an accelerometer embedded in loose clothes. Experiments are conducted using a wearable node embedded in an oblong piece of cloth to emulate loose clothes. The piece is attached to a participant's arm while performing daily activities. Measurements are collected and presented in both time and frequency domains. Finally, activity measurements are classified using SVM and KNN algorithms. Results indicate that the differences in measurements between loose and tight clothes are noticeable in both domains, but the degradation in classification accuracy is unneglectable. When the sensor was embedded in 10 cm long piece of cloth the classification accuracies are over 80% and 90% for SVM and KNN, respectively, which is approximate, 5% less than the tight clothes accuracies.