Implementation of a real-time text dependent speaker identification system

Speaker recognition has received for many years the attention of researchers working in the field of signal processing, but has not yet been implemented as a reliable feature in widely spread applications. We present in this paper a real-time text dependent speaker identification application, designed to consume a low amount of memory and processing power, serving as intermediary step towards the implementation as an embedded system. The speaker identification is based on computing the Mel Frequency Cepstral Coefficients and the derived Dynamic Coefficients, while classifying features using a Dynamic Time Warping approach. The software was tested using recording equipments similar to the ones integrated into mobile devices and shown satisfactory results.