On the Identification of FOSD-based Non-zero Onset Speech Dataset Conference Paper