The increasing use of relays in underwater acoustic (UWA) communications is a driving force to explore efficient techniques that can significantly improve the system performance. To fully exploit the benefits of cooperative systems, efficient management of resources including relays and power is required. In this paper, both optimal relay selection and power loading issues are investigated for an orthogonal frequency division multiplexing (OFDM) cooperative transmission over UWA channel. In our analysis, we consider amplify-and-forward (AF) relaying with perfect channel state information (CSI) at the source, relay, and destination nodes. Moreover, we assume sparse and frequency-selective Rician fading in the presence of colored Gaussian ambient noise. Unlike previous works on this topic, in our study, the power of noise is not identical for all subcarriers at both the relay and destination nodes. We solve two optimization problems that rely on the minimization of the bit error rate (BER) and maximization of the system capacity. In each problem, both optimal relay selection and power loading issues are addressed in two dependent phases. In the first phase, an unconstrained optimization problem is solved to determine the optimal relay out of multiple relays vertically located at different depths of water. We adopt all-subcarrier (AS) basis approach in our OFDM-based transmission model in which a single relay is engaged to transmit the entire OFDM block to the destination. In the second phase, after selecting the optimal relay, another optimization problem is solved to obtain the optimal power allocation. This is jointly done at both the source and relay nodes under total power constraint and fixed subcarrier rate. Extensive simulations are conducted to evaluate the performance of proposed algorithms under different scenarios.

Recognition of digital signal type is an important topic for various applications. In this paper a method is presented that identifies different types of digital signals. This method utilizes a radial basis function neural network as the classifier which an evolutionary algorithm, i.e. swarm intelligence (SI), is used to construct it. As the features of signals, this method uses fourth and sixth and eighth order of moments and cumulants, i.e. a combination of higher orders of statistics (HOS). In conjunction with neural network it is used K-fold cross validation to improve the generalization ability. Experimental results indicate that this method has high percentage of the correct classification to discriminate different types of signal even at low SNRs.