This research positively evaluated the vocality of the cough sound by estimating the average cough fundamental frequency or pitch. The fundamental frequency was determined by autocorrelation analysis on both the rough time signal and the linear predicted time signal [1]. An experimental cough database was recorded using a standard multi-media microphone (20 Hz - 20 kHz frequency response) and a sound card (16-bit) in a free acoustical field, using both male and female subjects. These included three ailing and nine healthy, non-smoking human subjects, along with two infected and two healthy Belgian Landrace piglets.

The healthy human subjects coughed voluntarily, whereas those suffering from a respiratory ailment coughed in an acute, productive way (i.e. sputum was cleared from their airways). Piglet coughing was either elicited by chemicals (in healthy animals) or was of a chronic, non-productive nature for subjects deliberately infected with bronchopneumonia. Accordingly, a total of 36 'voluntary', 48 'acute', 119 'chemical' and 1,883 'chronic' cough-sounds were studied.

Overall, 55% of 'voluntary', 75% of 'acute', 87% of 'chemical' and 73% of 'chronic' cough sounds were characterised by periodicity expressed by a global fundamental frequency. Generally, as in [2,3], the pitch of coughs originating from healthy subjects was higher than that of 'spontaneous' coughs resulting from ailment or infection. For both species, differences in pitch between healthy and infected subjects indicated the existence of acoustically different cough classes, in accordance with different causes or physical conditions of the respiratory system.

Pitch was also shown to vary during a single 'spontaneous' cough, so further research into the pitch-contour characterisation of the cough sound may be useful, as well as investigations into its production. The simplicity of free field sound recording and the objective, non-invasive nature of sound analysis offer great opportunities for implementating on-line follow-up work, and for applying this work as a diagnostic tool. The discovery of vocality represents an important and physically meaningful feature for advanced sound-recognition, consisting firstly of 'cough' versus 'no cough' recognition and secondly of whether a cough originates from a subject in 'pathological' versus 'normal' respiratory health.

References

1. Rabiner, L.R. (1977). On the use of autocorrelation analysis for pitch detection. IEEE Transactions: Acoustics, Speech & Signal Processing, 25(1), 24-33.
2. Korpas, J.; Sadlonona, J.; Vrabec, M. (1996). Analysis of the cough sound: an overview. Pulmonary Pharmacology, 9, 261-268.
3. Piirila, P.; Sovijarvi, A.R.A. (1995). Objective assessment of cough. Eur. Respir. J., 8, 1949-1956.

Paper presented at Measuring Behavior 2002 , 4^th International Conference on Methods and Techniques in Behavioral Research, 27-30 August 2002, Amsterdam, The Netherlands

© 2002 Noldus Information Technology bv