Myoelectric pattern recognition (MPR) based strategies are widely adopted in prostheses control to restore lost limb functions in amputees. In the MPR pipeline, high density (HD) surface electromyogram (sEMG) signals are acquired from a grid of electrodes placed on the amputees' specific residual arm muscles sites and then, processed to control the prosthetic device. Typically, electrode channels pick up sEMG signals from various muscles, which are then combined to decode individual targeted movements towards initiating requisite control for the device. Considering the significance of neural information across channels, especially in the case of transhumeral (TRH) amputees who often have limited residual arm muscles, we investigated the impact of two different electrode configurations (as placed on four different muscle groups) on limb movement intent classification. A total of four TRH patients were recruited, and HD-sEMG signals were acquired using a 32-channel recording system placed on different muscles of their residual limbs. The recorded signals were processed with a Wiener filter (WF), and a widely used time-domain feature with a linear discriminant analysis (LDA) classifier was implemented to decode the motion intents. From the experimental results, it can be concluded that the performance recorded by each muscle group is not dependent on the number of channels. Also, for both configurations, specific muscles play an important role in some tasks. It is anticipated that findings from this study would aid researchers in their experimental design for high-level upper limb amputees towards realizing a reliable and effective myoelectric PR control system.