Potential of self-conducted speech audiometry with smart speakers