Your english is totally fine, don't worry. As for the question, the problem is that you have to consider the alignment quality, base quality and the actual reference base to decide whether a read reliably indicates a variant at a given position. I would therefore probably run your 10 people through a variant calling pipeline, e.g. GATK (or at best the variant caller that was used to come up with this BED file) and then intersect the resulting VCF files with the BED file. Does that make sense in your case? In which format do you have your data? Something like BAM or fastq?