SEQMIX is a C++ program that takes advantage of off-targeted sequence reads from exome/targeted sequencing experiments for accurate local ancestry inference.
Before running SEQMIX, it is important to pre-process your data with a LD pruning step, which identify sites that are in high LD (r^2 > 0.1) and keep the sites with a higher sequence depth into the model. Since the sequence depth distribution is sample dependent, it is necessary to prune the sequence data for each individual.
Local ancestry inference with high density genotype array data can be done with existing software HAPMIX, LAMP, ANCESTRYMAP. (Warning: These programs are not user friendly and are very difficult to get it running)
Whole genome ancestry inference with ultra low coverage sequence data can be analyzed with LASER.