I can recommend the R package "spliceSites" to you and I'm not the developer.
It transforms Bam files into one big gap table, which basically is a large table with one row representing a gap in the alignment of the reads, which is represented by at least one read. Among others, it states the coordinates of the splice sites and the amount of reads, having the respective gap in their alignment. After annotating the gap table, it also tells you, if the splice donor or acceptor is annotated in the gtf file you use for the mapping and if not, it states the distance of the respective splice site to the next annotated splice site, which would be appropriate.
I'm studing splicing and after data preparation, I basically only work with these gap tables, since they describe quiet nicely, the read coverage of splice sites and the usage of not annotated ones.
You can also use multiple bam files to create one gap table, making it possible to compare the usage of a splice site across samples.
The read coverage of splice sites is stated in total reads and in rpmg, which represents the read coverage, normalized to the total read count in one bam file, or one bam file group.
I have to mention the 2-pass run of the STAR aligner, in case you don't know about it already, since it took me a while until i learned about it. Apparently, the 2-pass run with STAR is recommended by the STAR developer Alexander Dobin for analysis concerning splice site usage, esspecially anaylsis of not-annotated splice site usage.
So, if you are using R, I would definitly give the package "spliceSites" a try, in particular the function
readTabledBamGaps, since it gives you a very raw output, describing splice site usage in your data.
Edited some gramma