An Efficient Method for the Use of Overlapped Analysis Windows Employing the Goertzel Algorithm
Julian Penketh · 2002
For increased bandwidth savings in packet networks, speech compression algorithms are applied to reduce the amount of data that needs to be transferred to adequately reconstruct a telephone signal at the receiver. However, such algorithms do not process pure tones (such as dual-tone multi-frequency (DTMF)) with high fidelity, hence such in-band signalling tones must be detected before packetization occurs. To minimise the amount of DTMF tone distortion, and detect valid tones which may contain impairments, the tone recognition algorithm needs to decode tone-pairs in as short a time as possible. Many existing systems already use the Goertzel algorithm for DTMF tone detection. This paper presents a modification to the algorithm which allows an effective increase in detection frequency by overlapping, without the need for reprocessing of input data.