Embedded Files
Analysis of the evolution of the singing voice over the years
Music and Speech Modelling 2018
Abstract
Amongst the fans of the heavy metal band Metallica, James Hetfield's vocal transition after the ‘Justice’ album is often a topic of discussion. This project presents the evolution of his vocal characteristics throughout his career by analysing the track ‘Nothing Else Matters’ as recorded yearly from 1993 through 2018.
Introduction
Metallica is a heavy metal band formed in California in 1981. The band is lead by the singer James Hetfield, who throughout his career has evolved and adapted his voice. This project aims to investigate the evolution of the vocal characteristics of James Hetfield throughout his tenure as the singer for Metallica.
Aim
To study the transitions in the vocal characteristics of a singer during the live performances across the decades.
Motivation
Metallica, due to their worldwide popularity, was one of the earliest bands who exposed me to the world of Western Rock Music. Amongst the fans of Metallica, James Hetfield's vocal transition after the ‘Justice’ album is quite often a topic of discussion.
Material
Metallica’s ‘Nothing Else Matters’ being one of their softer numbers is also widely popular. I analysed various live performances of this track. The third verse and chorus are when James Hetfield stresses his vocals on the track. Hence I used the third verse and chorus as the corpus for this research. I segmented this section across the various live performance recordings and extracted his vocals track.
Method
A high-level procedure followed is as below
I obtain the live performance data via Youtube[1].
Convert the videos into '.wav' audio format.
The vocals are extracted/isolated from the rest of the audio.
I perform the following analysis
Pitch & Intensity
Fundamental (Formant) Frequency
Note Sustain Duration
Vibrato
- Vocal Extraction/Isolation Method
The third verse and chorus are segmented and loaded into Audacity. It was found that for most of the material the vocals were panned to the centre.
The vocals are first isolated using phase cancelling method wherein, the phase of the left channel is inverted and added to the right channel leading to the cancelling of the common signal between them.
This is now labelled as the 'Vocal Removed Track'.
A noise profile of the Vocal Removed Track is obtained via Audacity's inbuilt Noise Removal plug-in.
With this, the details of the non-vocal components of the signal are obtained. Next, the noise profile is phase inverted and cancelled out with the Original Audio to isolate the centre panned information.
The kick drum and the hi-hat are found to be also isolated along with the vocals as they are also mixed with centre panning. The vocals and drum signal is then passed through a band-pass filter with a low cutoff of 180Hz and high cutoff of 2000Hz to remove the kick drum and the hi-hat.
Resulting Audio
Unaltered Chorus Vocal Isolated Chorus
- Analysis Methods
Pitch & Intensity
From the 22 vocal isolated tracks, segment the word being sung on the beat.
In this analysis, the words Matters from the verse line And nothing else matters and Know from the chorus line And I know are analysed.
Pitch and intensity details are analysed using Praat.
The annual data gathered is compared with the same analysis of the studio recording.
The results are plotted using MATLAB
Figure 2: Pitch & Intensity Shift of the performances over the years when compared with the studio recording
- Discussion
From the top section of Figure 2, it is seen that in the early years (1995 - 2000) James Hetfield is unable to maintain the pitch. This is deduced to be due to the material in that era being performed under the effect of alcohol[2].
An answer to a question in Quora termed 'Why can't James Hetfield sing like how he used to?' quotes 'Hetfield’s singing style in the 80s and early 90s was very strenuous, singing very forcefully while also choking the air flow through the throat to create that gravelly, growling tone he was known for. That puts a lot of stress on the vocal cords, which can cause nodules and other damage if they aren’t properly cared for. Since that was the height of his drinking beer and wrecking shit on a daily basis phase, he probably wasn’t caring for his voice like he should have, and he did end up blowing out his voice. Recovering from that almost always involves adapting your technique, so you never sound quite the same.' which further supports this results.
During the second half of the analysis years, the shift in pitch is found to be in correlation with the National Centre for Voice and Speech[3]. The result also indicates signs of Ossification (hardening) of the laryngeal cartilages.
Fundamental Frequency
From the 22 vocal isolated tracks, segment the word being sung on the beat.
In this analysis, the words Matters from the verse line And nothing else matters and Know from the chorus line And I know are analysed.
Formants are detected using Tony (tool for melody transcription).
For certain materials, the annotations are octave shifted and adjusted manually when the in-built predictions were inaccurate.
The annual data gathered is compared with the same analysis of the studio recording.
The results are plotted using MATLAB
Figure 3: Fundamental Frequency data for the word 'Matters' in the verse, consolidated over the years.
Figure 4: Fundamental Frequency data for the word 'Know' in the chorus, consolidated over the years.
Figure 5: Pitch Shift of the performances over the years, when compared with the studio recording, analysed using data obtained via Tony
- Discussion
Analysing Figure 3, it is observed that there is a transition in the pitch over the years. It is seen that the material can be segregated into different classes based on the shapes of the formants.
Years 1992 - 1996 have a flatter ending of the note.
Years 1997 - 2000 have two troughs towards the end.
Years 2004 - 2015 return to the flatter ending of the note.
Years 2015 - 2018 introduce vibrato towards the end.
Similar to the results of 3, analysing Figure 4 reveals a similar trend in the transition in pitch over the years. It is observed in Figure 4, that the duration of the word seems to diminish and the roll-off in the curve increases as the years progress.
Figure 5 indicates a similar analysis performed earlier using Praat, but this time with the data obtained via Tony. The data obtained via Tony, however, provide the exact result obtained using Praat.
Because of the results obtained in this analysis, the next two analysis was conducted.
Note Sustain Duration
From the 22 vocal isolated tracks, segment the word being sung on the beat.
In this analysis, the words Matters from the verse line And nothing else matters and Know from the chorus line And I know are analysed.
The length of the words sustained is marked by measuring the start time and the end time of the word sung in the audio track via Audacity.
The results are plotted using MATLAB
Figure 6: Note Sustain Duration of the performances over the years.
- Discussion
Figure 6 shows the note sustain duration through the years. It can be seen that the sustain duration declines as the years' progress.
Vibrato
From the 22 vocal isolated tracks, segment the word being sung on the beat.
In this analysis, the words Matters from the verse line And nothing else matters and Know from the chorus line And I know are analysed.
Formants are detected for the words which are under the effect of vibrato using Tony.
For certain materials, the annotations are octave shifted and adjusted manually when the in-built predictions were inaccurate.
The audio along with the pitch data is exported to Sonic Visualiser.
The oscillation in the formant frequency is analysed using the spectrogram in Sonic Visualiser.
The results are plotted using MATLAB
Figure 7: Vibrato oscillations observed in the performances over the years.
- Discussion
Figure 6 shows the vibrato observed in the performances through the years. The effect of the vibrato is observed from 2013. Post-2013, it can be seen that the frequency oscillation has increased from +/- 2Hz to +/- 5Hz in 2018.
According to the National Centre for Voice and Speech, the presence of vibrato in the voice is another indicator for Ossification (hardening) of the laryngeal cartilages[6]. On this occasion, it could not be confirmed whether the appearance of vibrato is an indication of the singers' expressivity or occurrence of the aforesaid physical transformation.
Conclusion
Throughout the career of James Hetfield, his voice
Reduced in pitch and then stabilised.
Decreased note sustain duration.
Increased signs of vibrato.
Indicates Ossification (hardening) of the laryngeal cartilages.
The transition in his vocal characteristics is in correlation with the singers' lifestyle. (during the late 90s and early 2000s)
Further Work
Perform a listening survey and obtain the listeners preferred voice over the years.
Analyse the correlation in the transitions obtained in the vocal characteristics to the regular speech of the singer.
References
Youtube Resources
During the recording of the band's eighth studio album St. Anger in 2001, Hetfield went into rehab to address his alcohol addiction. He rejoined the band after seven months in rehab and four months recovering with his family. https://en.wikipedia.org/wiki/James_Hetfield
National Center for Voice and Speech - Voice Changes Throughout Life, "http://www.ncvs.org/ncvs/tutorials/voiceprod/tutorial/changes.html"
Mysteries of the Ageing Voice “http://www.vulture.com/2016/10/mysteries-of-the-aging-voice.html"
What Happened to Bob Dylan’s Voice? A Doctor Explains “http://www.vulture.com/2015/02/why-does-bob-dylan-voice-sound-like-that.html"
More Bone in Your Tone: The Consequences of Laryngeal Ossification, "http://www.claudiafriedlander.com/the-liberated-voice/2014/07/bone-tone.html"
Singing voice analysis/synthesis “https://dspace.mit.edu/handle/1721.1/62044"
Singing voice analysis/synthesis “https://dspace.mit.edu/handle/1721.1/62044"
Changes in the male voice at puberty “http://adc.bmj.com/content/77/5/445.short"
The Acoustics of the Singing Voice “http://www.jstor.org/stable/24953939?seq=1#page_scan_tab_contents"
The Science of Singing Voice “http://asa.scitation.org/doi/pdf/10.1121/1.399243"
Old voices: What do we really know about them? “https://www.sciencedirect.com/science/article/pii/S0892199787800188"
Acknowledgements
I would like to acknowledge Professor Elaine Chew and my colleagues on the Music & Speech Modelling Module who have helped me with this study by providing constant feedback.
Page updated
Google Sites
Report abuse