The isunicode function may look promissing, but doesn't actually work on my test files (generated with Notepad++). The same problem occurred with Walter's detect_UTF_encoding function (for which I appear to have misplaced the thread it was posted in). Trying to detect the encoding is a tricky business.
But if you want to remove any Unicode chars you can also use my readfile FEX submission. It will try to determine if a file contains UTF8 chars and will handle the conversion. You should be able to use this function to read your file and discard any >128 chars.
Note that if you're using Matlab 6.5 or GNU Octave, some restrictions apply to what files are read succesfully.