Spatial Modulation (SM) is a promising multi-antenna technology proposed for 5G networks due to its numerous features, including high-Energy Efficiency (EE) and low transceiver complexity. A variant of SM is Space Shift Keying (SSK) that further reduces detection complexity at the data rate cost. However, SM and SSK forms have limitations that yield Bit Error Rate (BER) degradation and low Spectral Efficiency (SE). Transmit Precoding (TP), and Non-Orthogonal Multiple Access (NOMA) are well-known solutions to these problems. Nevertheless, many existing TP algorithms and combination models between NOMA and SM/SSK (NOMA-SM/SSK) have shown several restrictions that led to high complexity and impracticality. Therefore, this thesis aims to enhance the BER and SE of SM and SSK and solidify the practical feasibility of TP and NOMA-SM/SSK. The TP approaches are classified into Codebook-based Precoding (CP), Non-CP (NCP), and Hybrid-CP (HCP). CP avoids the necessity for a complete knowledge of channel state information and relaxes processing complexity in the NCP and HCP. NOMA-SM/SSK models are categorized into Multi- and Single-Radio Frequency (MRF/SRF) models based on their needs for RF chains. In contrast to the MRF model, SRF preserves EE and low complexity SM's features. Four aspects are tackled in this research. Firstly, the CP approach minimizes the BER of the single-user SSK system by designing a full-combination codebook based on phase-rotation precoding with a systematic and permanent structure composed of only four phases. A new closed-form BER for non-precoded SSK is also derived. Secondly, the feedback overhead and receiver's processing time of the CP approach are optimized via presenting a new codeword selection criterion and low complexity codebooks. Thirdly, an SRF Generalized (SRGen) is proposed as an efficient NOMA-SM/SSK model to improve the SE of the multi-user SM/SSK system. The users jointly decide their shared antenna and separately select their symbols in the SRFGen. It creates several SRF special cases, including the existing SRF models, and overcomes their limits. Lastly, SRFGen and MRF models are extended for an arbitrary number of users, and their SE is analyzed for Gaussian and finite-alphabet inputs. New antenna detection and power allocation methods are also introduced. Numerical results show that the designed CP and SRFGen NOMA-SM model respectively enhance BER and SE of SM with more suitability for implementation than their available counterparts. The designed CP significantly minimizes the BER than the non-precoded SSK (5.5 dB Signal to Noise Ratio (SNR) gain at 4-phases) and offers a comparable BER (1 dB difference) and lower complexity (21% reduction) than the extant NCP. The outcomes also reveal that the SRFGen delivers high EE (more than 48%), and robust SE competitor to the MRF model (less than 0.2 bits/s/Hz difference), particularly for large numbers of antennas and users. Sharing the antenna selection in the SRFGen and its special case SRFC2-L having L SSK signals offer higher SE (0.2 bits/s/Hz gain), lower BER (about 1 dB gain), and efficiently utilize the spatial domain than the existing SRF models (i.e., SRFC1 and SRFC2-1-UnShared).
