Single-lead electrocardiogram (ECG) from wearables is prone to motion artifacts, caused by day-to-day human activities, that makes it difficult to obtain relevant information such as identifying QRS complex and computing R-R intervals for clinical and wellness applications. This issue requires the development of complex signal processing techniques or the use of multi-sensor fusion modalities to eliminate the noisy motion artifact recordings from use during analysis. However, these approaches require the use of other sensors such as inertial measurement units (IMU) which in turn increase the power needs of small battery-powered singe-lead ECG devices. This work proposes an exploration of embedded machine learning (ML) models in the classification of capacitive electrode-based ECGs that contain motion artifacts in the form of different types of human movements. The supervised ML models investigated are ensemble-based methods, support vector machines (SVM), naïve Bayes, and k-nearest neighbors (KNN) with 10-fold cross validation. Capacitive-electrode based ECGs are obtained from five adults under different upper-limb movements including asymmetrical actions such as punching, running, body twisting, symmetrical actions such synchronized arms flapping, and no action. A total of 48 statistical features are extracted from the original ECG signals and its wavelet coefficients based on Daubechies wavelet 4, db4, decomposition. The ensemble-based KNN approach is shown to provide the highest accuracy of 92% and sensitivity of 95% in binary classification between ECGs containing any of the four different movements and ECGs with no activity. However, SVM showed the highest accuracy of 68% in multi-label classification between asymmetrical, symmetrical, and no action movements compared to 64% for KNN. This work is a preliminary step in the investigation of embedded and explainable AI (xAI) techniques in identifying different activities solely from the motion artifact noise information presented in capacitive electrode-based ECGs used in wearable devices.