Some experimental methods such as TAOS [S. Holler et al., Appl. Optics 43 (33) (2004) 6198-6206] are capable of collecting the light scattered by single airborne particles in the micrometer size range when the latter are illuminated by a triggered LASER source. Data consist of intensity patterns collected in a suitable solid angle and made available at a high rate (\>100 patterns per second). There is no known theoretical method capable of determining the particle size, shape and complex refractive index from such data. As a consequence a heuristic algorithm has been developed, which relies on spectrum enhancement for feature extraction and on multivariate statistics for classification. Spectrum enhancement of an image amounts to the application of a pseudo-differential operator followed by a nonlinear transformation with the aim of separating structure from texture. The classifier has been trained with patterns from two materials (polystyrene spheres and highly irregular elastomer particles). Training has involved the solution of saddle point problems and multiobjective optimization with respect to the parameters which control the algorithm. Recognition has been applied to patterns in the data set originated from different materials. The location, in the principal components plane, of the pattern yielded by a given particle has been related to deviation from the spherical shape, in agreement with findings from scanning electron microscopy. Results show that the classifier has acceptable performance in terms of error rate and has real-time potential, hence is applicable to environmental monitoring.