The distance set representation of speech segments