A new cosine series antialiasing function and its application to aliasing-free glottal source models for speech and singing synthesis.