Speech Recognition Approach From Midi File

Abstract

In this paper we explore a new technique for speech recognition from recorded human speech on midi file .The microphone is used to record the statement in the computer .The speech features are extracted from input track using a proposed algorithm corresponding to the melody to be used in the recognition algorithm. The aim is to build a reliable and efficient large scale system that collect thousands of sound as a Midi data, finally the system will returns list of the wave files most similar to the humming and recognize the input sound. In this paper we record one statement in available language of 200 different person from our city (alnassiryha in south of Iraq) and store as a system data base. Promising result were obtained testing different Midi files as experiments.