Electromyogram (EMG) has emerged as a promising technique for grasping force (GF) estimation because of its non-invasive nature of measuring muscle activity. GF estimation plays a crucial role in developing prosthetic devices and rehabilitation strategies for individuals with upper-limb impairments. Research established the need for prosthetic hands with the ability to control GF like human hands. However, commercial versions of existing prosthetic hands lack GF control mechanisms. This indicates the need for more research on prosthetic hands for appropriate GF control. The work presented in this manuscript investigates the estimation of GF based on EMG using polynomial regression and k-nearest neighbor (KNN) models. Two channel EMG were acquired from the flexor digitorum profundus and extensor digitorum muscles of the right upper limb from six healthy subjects during grasping. An electronic hand dynamometer (EH101) was used to collect force data while the dynamometer was being pulled during grasping. The polynomial regression and KNN models had mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE) of 1.72 ± 0.016, 1.33 ± 0.037, 1.016 ± 0.004 and 0.74 ± 0.022, 0.86 ± 0.018, and 0.21 ± 0.005, respectively. Additionally, R2-values of the estimated GF and actual GF were 0.72 ± 0.039 and 0.88 ± 0.017, respectively. The proposed methods provide an effective solution to estimate GF based on EMG, contributing to the better controllability of EMG controlled prosthetic hands.