today, the digital terrestrial television (DTT) market is characterised by the high capacity required to transmit high-definition TV services and the available spectrum. Efficient use of radio spectrum is therefore needed, which requires new technologies to ensure increased capabilities. Non-uniform constellations (NUC) emerge as one of the most innovative techniques to address such requirements. NUCs reduce the space between the QAM uniform constellations and the theoretical Shannon boundary. With these constellations, the symbols are optimised in both phases (I) and square (Q) by geometrical signal modelling techniques, considering a specific Noise Signal Level (SNR) and a specific channel model. There are two types of NUC, one-dimensional and two-dimensional (1D-NUC and 2D-NUC respectively). 1D-NUC maintains the square form of QAM but allows changing the distribution between symbols in a particular component, with a non-uniform distance between them. These constellations provide better SNR performance than QAM, without any increase in complexity in demapper. The 2D-NUC also makes it possible to change the square shape of the constellation, making it possible to optimise the symbols in both dimensions and thus obtain greater gains in capacity and lower requirements in SNR. However, the use of 2D-NUCs implies greater complexity in the receiver. In this thesis, NUCs are analysed in terms of both transmission and reception, using either antenna configurations (SISO) or multiple antenna configurations (MIMO). In SISO transmissions, 1D-NUCs have been optimised for a wide range of different SNR and several constellation orders. The optimisation of 2D-NUCs roasted has also been investigated. Although complexity does not increase, the SNR gain from these constellations is not significant. The highest turnover gain is obtained for low constellation orders and high SNR. However, using multi-RF techniques, the gain increases dramatically as components I and Q are transmitted on different RF channels. In this thesis, several multi-RF gains representative of NUCs, with or without rotation, have been studied. At the recipient, two different bottlenecks have been identified in the implementation. First, the complexity in the receiver has been analysed for all the constellations considered and, subsequently, two algorithms are proposed to reduce complexity with 2D-NUCs. In addition, the two can be combined into a single demapper. The quantification of these constellations has also been explored, as both the LLR values and the I/Q components are modified compared to traditional QAM constellations. In addition, an algorithm has been proposed that is based on optimisation for different levels of quantification, for a specific NUC. The use of NUCs in MIMO has also been investigated in detail. Optimisation in one or two antennas, the use of power imbalance, factors of discrimination between receiving antennas (XPD), or the use of different demappers have been included. Assuming different values, new multi-antenna constellations (MA-NUC) have been obtained thanks to a new re-optimisation process specific to MIMO. In the recipient, the complexity analysis has been extended in demapper, which is greatly increased by the use of 2D-NUCs and MIMO systems. Alternatively, a solution based on the Soft-Fixed Sphere Decoding algorithm (SFSD) is proposed. The main problem is that these demappers do not operate on 2D-NUCs, as they need an additional step in which components I and Q need to be separated. The proposed method quantifies the nearest symbol using the Voronoi regions, allowing the use of this type of receptor.