Analysis of Tahreer in traditional Iranian singing

The song of nightingale is regarded as the symbol of musical beauty in Iranian visual arts, literature and poetry. Persian/Iranian traditional singing, āvāz, is often accompanied by a vocal ornamentation called Tahreer which is perhaps inspired by the singing of nightingale too. Tahreer is a quick alternation between laryngeal mechanisms M1-M2-M1 (chest-Falsetto-Chest), producing a frequency jump during a very short time interval, typically 50 to 70 ms. Due to the fast nature of the technique, it is not perceived as a transition but rather an abrupt breaks in a continuous melody (in M1).  Tahreer could be called the Iranian Yodelling with the difference that unlike a yodel melody expanding successively in M1 and M2, Tahreer melody stays completely in M1 with short ornamental excursions in M2. What follows is an analysis of different features of Tahreer to help expose more information about its characteristics.

                                                                                                                                                Spectrogram of a segment of nightingale song (left) and Tahreer (right)

Dataset

There are a number of contemporary schools of avaz which differ mainly by the style of ornamentations. A selection of renowned singers were chosen from different schools of avaz as well as different eras (1920s - present) to encompass different styles and act as a good sample set. Different vocal segments from each singer were manually chosen to ensure a good presence of Tahreer and minimal instrumental accompaniment for clearer analysis. 

Ghamarolmolouk Vaziri(20s-50s)              Jalal Taj Esfahani (30s-60s)                   Mohammadreza Shajarian(60s-present)    Alireza Eftekhari (70s-present)          Alireza Ghorbani(90s-present)

Is it a Vibrato?

 looking at the spectrogram of a Tahreer at the first glance it looks similar to a Vibrato due to the visible oscillation in frequency. There are however major characteristic differences between Tahreer and Vibrato still visible by just looking at the spectrogram. Tahreer has larger deviation from the main note, sharper rise and fall and a shorter duration. Tahreer could also have asymmetrical rise and fall, with the rise being faster and the fall more gradual. Below there are 4 instances of Vibrato (Row 1) and 4 instances of Tahreer corresponding to the same performance (row 2).

Jump Duration

In order to calculate the duration of the short jump to the falsetto note in the performance of each Tahreer a number of short segments containing Tahreer were selected from the corpus. For each segment the start and end of the jump were manually annotated using sonic visualizer as depicted below. The start point was selected as the time when the melody line starts ascending and the end point , the time when the descend is completed. 

This data was then exported and analysed in MATLAB to calculate the duration of each instance of Tahreer (purple bands above) for all selected segments.

Mean and Standard deviation were calculated for all Tahreer instances per singer. The table below contains the results ranging between 57 and 85 ms with a standard deviation of 10-19 ms. The longest mean and highest standard deviation both belong to one singer. By disregarding the corpus related to the singer the Mean duration of all other singers is roughly 60 ms with standard deviation of 10 -12 ms.

Jump distance

Due to the speed of the transition the higher (secondary) note is not audible in normal playback speed. So an interesting analysis was to determine the pitch of the secondary note and determine if a relationship exists with the primary note. Using the software Tony , the melody line was transcribed and the data exported to MATLAB for analysis. The segments chosen were again specifically selected to contain only Tahreer and therefore using peak picking the local minima and maxima were chosen which would be the primary and secondary notes respectively. 

Four segments containing continuous instances of Tahreer from different performances were selected per singers. The mean distance were calculated for each singer as well as the overall mean for all singers. In order to have a more musically meaningful result , frequency differences were changed to estimated note intervals prior to calculating the means.

A few observations were made:

1. If a primary note is repeated , the secondary notes following each would stay relatively constant too. 

2. The assumption above does not hold on the last primary note of a series. If the melody is descending following the last note , the secondary would be of a lower pitch and if the melody is about to ascend, the secondary would be of a higher pitch.

3. The relative distance of the secondary note from the primary was estimated to be at least 2 semitones and at most 5. Except for one singer who used 4-5 semitones as the distance , the average duration would be 2-3 semitones for all singers.

   1. The jump measurements recorded contain the whole range between 2 and 3 semitones which suggests a rough attempt to have a jump of at least a major second and at most a minor third but with no necessary obligation to ensure landing on whole semitones. No further deductions were possible at this level of analysis without considering other parameters such as scale, melodic motif or primary pitch.

Change in loudness

During the performance of Tahreer a Tremolo type of effect can be heard which is also easily identifiable by looking at the time-domain representation of a recording. The image below (left) shows that the intersection between each segment (of amplitude) roughly corresponds to the peak point of each Tahreer in the spectrogram. This is further supported by the raw power overlaid using the green dotted lines below(right) that depict the local minima in amplitude at the point of each jump in frequency. This may may due to both the short amount of time spent on the secondary note and the change to the higher vocal mechanism which could possibly result in quieter output.

Aspiration using the phoneme "h"

In majority of cases every instance of Tahreer is accompanied with aspiration when moving from the primary to the secondary note. This results in the utterance of an "h" sound which is the most audible part of this technique in normal speed playback. For instance the word "Jaan" which is a popular lyrical phrase used for adding Tahreer at the ending of poem verse, would become "Jaahaahaahaahaan" after using 5 instances of Tahreer. It is not easily possible to observe the added aspiration in time or frequency domain due to the voiceless nature of the sound but it is easily audible and to the listener in fact perceived as an intentional insertion of an "h" sound rather than the aspiration used for the transition. Untrained signers imitating the Tahreer would often start by trying to insert "h" sounds in the middle of long vowels while singing.

Other regional vocal ornamentations

 Other styles of Iranian singing contain similar techniques to a Tahreer for adding the ornamental element. Arabesque music of the 50s and 60s inspired by famous Arab singers such as Umm Kulthum as well as combinational styles of pop singing developed in the second half of the 20th century use a considerable amount of their own style of Tahreer. All these Tahreers add ornamentation to melodic content, especially when moving between notes and to an unfamiliar ear they could sound similar to the traditional Tahreer but their key difference is the lack of jumps to a secondary note. An example of this could be seen in the image below (left) of a vocal segment of a pop song. It is possible to see the short breaks in the melody but the transition to a higher note is a lot less visible. The image on the right is of the arabesque style where the jumps to higher notes are actually prolonged audible visits, making them part of the melody. 

Arabic Tahreer style which inspired the Iranian Arabesque depicted above, is also made of audible alterations with neighbouring notes of the scale resulting in a smooth melodic transition. Below is a short excerpt from a prominent Egyptian singer, Mohammed Abdel Wahab, singing and using Tahreer and performing 8 notes in a short time (1.4 s).

The form of  Tahreer discussed in this project is unique to Iranian , Kurdish and Azeri music and interestingly the Tahreer used in neighbouring Iraqi and Turkish music is a lot closer to the above example and the wider middle eastern style of singing. Further research, literature review and  analysis is required to be able to get a solid picture and be able to explain the conforming and contrasting elements of the techniques used in these neighbouring cultures.

Discussion

This project aimed to depict a better picture of a unique vocal technique which is perhaps less known and researched to date. Traditional Iranian music is an ancient art form which has been and still is taught via a direct masters to student relationship through generations. Music notation and available literature so far have not been able to capture the depth and subtleties of the techniques enough to allow for more accessible methods of proper learning and for the case of singing it is even harder to refer to written text for any form of learning. By utilising computational techniques and analysing different characteristics of Tahreer it is possible to learn more about the subtleties and hidden features of it, such as the short jump and transition discussed earlier. These findings coupled with more extensive research in the future could allow for new forms of transferable knowledge to be created and provide new and more accessible platforms for future students to learn the techniques and ornamentations to some extent. 

Discussing detailed characteristics of Tahreer and comparing it to other styles of vocal ornamentation would also allow for ethnomusicology researchers and generally people interested in music of the Middle east to have a better understanding of the variety, contrast and depth of different musical techniques used in each region, culture or country.

Bibliography

Caton, M. (1974). The Vocal Ornament Takiyah in Persian Music. UCLA Selected Reports in Ethnomusicology.

Michèle Castellengo, J. D. The Iranian tahrir: Acoustical analysis of an ornamental vocal technique. Cim07.

Biglari, H. J. Timbral and melodic characteristics of the Persian singing.

Bernard Roubeau, N. H. (2007, Oct). Laryngeal Vibratory Mechanisms.

Nathalie Henrich, M. K. (2006). Resonance tuning and glottal behaviour in different vocal qualities in Bulgarian women's singing.