Modeling distortion product otoacoustic emission input/output functions using segmented regression.