The thesis is concerned with the estimation of the sampled impulse response of a time varying HF channel, where the estimators are used in the receiver of 4800 bita/s quadrature amplitude modulation (QAM) system, operating at 2400 bauds. A number of channel estimation techniques are tested using an idealized model of the data transmission system. Estimation techniques studied are the simple gradient estimator, the adaptive channel estimators, the improved channel estimator (that takes into account prior knowledge of the number of fading paths in the channel) and the neural network-based channel estimator. These estimators are tested over three different HF channel conditions. The results of these tests have shown that the adaptive channel estimators have performances intermediate between those of the simple and improved estimators but are only a little more complex than the simple estimator. The tested neural network-based estimator has a performance slightly better than the simple estimator. Channel estimation may also be achieved by estimating the values of the channel parameters (parametric estimation), rather than estimating the channel impulse response. Such estimation scheme is suitable when the link quality is to be qualified in adaptive communication systems for communication system parameter adaptation. In this thesis, estimators for the signal-to-noise ratio and the Doppler spread are proposed together with a measurement technique for the severity of the channel distortion. Computer simulation test results over different channel conditions have shown that these channel parameter estimators are efficient and valuable.