Formant frequency adjustment in barbershop quartet singing

In this master thesis, the formant frequencies of an excellent male Swedish barbershop quartet were examined. The choice of topic was based on several factors that ought to make the tuning of the formant frequencies extremely important in the strive for the characteristic barbershop sound. The quartet was recorded, singing parts of the song 'Paper Moon'. In the recordings, each singer had an individual microphone taped on his nose, and apart from recordings when the quartet sang altogether, each singer was also recorded alone, singing his voice (part), and reading the lyrics. All recordings were made three times. The examined material, that turned out to be used, was the first four formants of the vowels, 'to' (/u/), 'be' (/i/), and 'divine' (/a/), and the progression of the formants in the extended fermata, that ended the song (referred to as the song's 'tag'). The determination of the formant frequencies was made by filtering of a short time interval of the sound file of interest, with filters that corresponded to the inverses of the formants. The formant frequencies could then be determined with good accuracy, by tuning the formants' frequency and bandwidth so that the inverse filtered signal looked like a realistic waveform of the transglottal airflow. Some problems and sources of error in the study were: • The falsetto singing, performed by the tenor. This made the shape of the transglottal airflow waveform anonymous and, hence, impossible to identify, when carrying out the inverse filtering. The tenor's formants could therefore only be determined for the vowel 'to' (which was not sung in falsetto). • Nasality. The transglottal airflow waveform became buckled when the singers sang in a nasal manner. This deteriorated the accuracy of the determination of the formant frequencies. • Leakage of sound from the singers' to the other singers' microphones (crosstalk). In the vowels 'be', and 'divine' the baritone's recording was affected by leakage, which again deteriorated the accuracy of the determination of the formant frequencies. In the tag, the bass's recording was affected by leakage. 3 The accuracy in the determination of the formant frequencies had an approximate standard deviation of 2.2 % of the measured value. The results of the study strongly indicated that the singers strived to separate their own formants from the others. Since the intonation in barbershop singing is of extreme importance, it is likely that the singers spread their formants in order to …