Abstract
The use of meteor ionisation trails as ‘cheap satellites’ to reflect radio waves between two points on the earth's surface is an established technique, called Meteor Burst Communications (MBC). For MBC systems to take advantage of the different amplitude and duration patterns of different trail types it is necessary to predict these patterns from features of initial signals reflected from the trails. The work described in this paper attempts to predict trail amplitude, duration, and trail type using neural networks. Results include a picture of what features of the beginning of the trail are most and least important for recognising various characteristics of the rest of the trail, some significant results as regards trail type prediction, and high correlations between actual and predicted peak amplitudes of trails. The latter is an important result.