ICSV22, Florence (Italy) 12-16 July 2015 1 PERFORMANCE COMPARISON OF DIFFERENT CONTACT MICROPHONES USED FOR VOICE MONITORING Alessio Carullo, Federico Casassa and Antonella Castellana Department of Electronics and Telecommunications, Politecnico di Torino, Corso Duca degli Abruzzi, 24 - 10129 Torino, Italy e-mail: antonella.castellana@polito.it Arianna Astolfi, Lorenzo Pavese and Giuseppina Emma Puglisi Department of Energy, Politecnico di Torino, Corso Duca degli Abruzzi, 24 - 10129 Torino, Italy The aim of this paper consists in comparing the performances of different sensors used as contact microphones for voice monitoring. Four devices are analyzed: two electret condenser microphones which differ in size, an accelerometer and a piezoelectric transducer. The inves- tigation is based on the analysis of the signal acquired at the output of each sensor when it is attached at the jugular notch, to sense the skin vibration induced by the vocal folds activity. This signal is often used to extract vocal parameters, such as sound pressure level, fundamental frequency, time dose, jitter and shimmer, which are of interest for diagnostic purposes as well as for vocal load assessment. Experimental tests were performed in a sound absorbing chamber using different speech materials. During these tests, a calibrated microphone placed at the fixed distance of 17 cm from the speaker’s mouth was used to ensure measurement traceabil- ity. The signals simultaneously acquired at the output of each contact sensor and of the micro- phone in front of the speaker’s mouth were compared to identify which device better provides exhaustive information of the speech signal. Since the use of these transducers is not intrusive and does not interfere with normal daily activities (e.g. for voice long-term monitoring in working environments), their sensitivity to body movements and to background noise has also been estimated. The frequency response of each sensor compared to one of the signal acquired by the microphone in air was furtherly analyzed. 1. Introduction Voice disorders at different levels, from simple vocal dysphonia to vocal fold nodules [1], are in- creasingly affecting those professional categories that make use of voice in a sustained way and for prolonged periods of time (e.g. actors, singers, call-center employees, sales people). The use of voice in occupational settings causes significant changes in vocal parameters with respect to non- working days [2], and the appearance of voice disorders may bring to absenteeism from work in order to recover [3], although all the related causes are still in course of study [1], [4], [5]. Long-term monitoring of voice are therefore needed to characterize the vocal behavior of voice professionals during working activities, as well as to identify the degree of dysphonia. Several works have already dealt with the issue of monitoring the speech signal [6], [7] and different devices for vocal activity monitoring have been developed. Recently, the interest towards