There is disclosed a
system for inputting speech that comprises an echo canceller section 120 for subtracting a pseudo
echo signal, generated by an
adaptive filter 121 based on a speech
signal inputted to a first input terminal 120a, from a speech
signal for sending as inputted to a second input terminal 120b to thereby remove echo components contained in the speech
signal for sending; a main
microphone 151 for mainly receiving sound from an objective sound source; a reference
microphone 152 disposed a predetermined distance apart from the main
microphone 151 for mainly receiving sound from other than the objective sound source; and an audio
delay section 180 for giving a predetermine
delay to the speech signal for sending from the main microphone 151. The speech signal from the reference microphone 152 is inputted to the first input terminal 120a of the echo canceller section 120, while the speech signal for sending which is from the main microphone 151 and to which the
delay has been given in the audio delay section 180 is inputted to the second input terminal 120b of the echo canceller section 120. The delay at the audio delay section 180 is adjusted within a range of
delay time of the echo components which the echo canceller section 120 removes. According to this
speech input system, the
noise is well removed regardless of directions of the sources of the objective sound and the
noise, regardless of the difference in level between the objective sound and the
noise and / or regardless of whether the noise abruptly changes its tone or volume, so that a clear
speech input is achieved even in such a condition that there is noise in the surrounding environment.