Acoustic Description of Bhramari Pranayama

Abstract

The study's aim was (1) To describe the acoustic characteristics of Bhramari pranayama, and (2) to compare the acoustic features of nasal consonant /m/ and the sound of Bhramari pranayama produced by yoga trainers. Cross-sectional study design. Thirty-three adult male yoga trainers performed five repeats of nasal consonant /m/ and Bhramari pranayama. These samples were recorded into Computerized Speech Lab, Kay Pentax model 4500b using a microphone (SM48). Formant frequencies (f_F1, f_F2, f_F3, & f_F4), formant bandwidths (B_F1, B_F2, B_F3, & B_F4), anti-formant, alpha and beta ratio were analyzed. Nasal consonant /m/ had higher f_F2 and anti-formant compared to Bhramari pranayama. Statistical significant differences were noted in f_F2, B_F3, and anti-formants. Bhramari pranayama revealed a low alpha ratio and a higher beta ratio than /m/. However, these differences were not statistically significant. Findings are discussed from acoustic and physiological perspectives. Bhramari pranayama was assumed to be produced with a larger pharyngeal cavity and narrower velar passage when compared to nasal consonant /m/. Verification at the level of the glottis and with aerodynamic parameters may ascertain the above propositions.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12070-021-03054-1.

Introduction

Pranayama is the formal practice of controlling the breath for a healthy body and mind [1]. The word prana means the vital energy for life, whereas Ayama represents ‘control’. Hence pranayama translates to ‘control of the vital energy, i.e., breath for a healthy life'. Pranayama is the fourth limb of Ashtanga yoga that deals with exercises related to the breathing pattern. Regular practice of pranayama is claimed to relieve stress and anxiety, correct the hormonal imbalance, improve lung-related diseases and other voice disorders [2]. Bhramari pranayama is an ancient ‘yogic’ breathing exercise that includes a unique breathing technique, performed in a relaxed position with simultaneous generation of a constant humming sound during expiration [3]. In Bhramari pranayama, one should sit in a comfortable posture and inhale & exhale through nostrils slowly and deeply. While exhaling, a sound (humming sound) like a bumblebee will be produced through the nasal cavity, keeping the oral cavity closed by the lips and ears closed by fingers [2] (Demonstration of Bhramari pranayama: https://youtu.be/vA_Ib_QAWW8). Bhramari pranayama is thought to expedite the purification of pranic channels and balance the flow of breath [1–4]. Humming in Bhramari pranayama facilitates the easy onset of phonation [5] and resonant voice [5, 6]. Bhramari pranayama is claimed to produce a positive effect on voice quality [6].

Resonant voice is defined as a voice produced with ease and vibration in facial tissue and is considered a target for a healthy voice [7]. Nasal sounds are produced with a closed oral cavity and open velopharyngeal port⁸. The closed oral cavity has a complete obstruction to airflow in the oral tract, and thus nasal sounds such as /m/, /n/, and /η/ are termed as ‘nasal consonants’ [8]. The interval in which the oral closure coincides with an open velopharyngeal port is referred to as the nasal murmur [9]. Acoustic characteristics of nasals are (a) Low first formant frequency about 300 Hz (nasal formant), (b) narrow spacing between formants reflecting the dampening of energy (large bandwidth reflecting a rapid rate of sound absorption), and (c) the presence of antiformants [10].

Voice and its resonance can be described acoustically. The techniques are broadly grouped into the measure of periodicity [Pitch, Amplitude, Harmonics to Noise Ratio (HNR), Spectral & Cepstral], intensity (Phonetogram & Voice Range Profile), and spectral shape [Long Term Average Speech Spectra (LTAS)] [11]. In addition to formant frequencies and bandwidth, LTAS documents the source and formant characteristics of speech sounds [12]. Measures of LTAS such as intensity values of alpha, beta, and gamma ratio demonstrate the amount of energy present at each frequency pooled across time [12]. Alpha, beta, and gamma ratios are calculated by the formula given below [12], where ‘E’ stands for energy measured as the amplitude of the signal in the specified frequency region.

• α ratio: E1/ E2—where E1 is energy in 0–1 kHz and E2 is energy in the 1–5 kHz region

• β ratio: E3/E4—where E3 is energy in 0–2 kHz and E4 is the energy in 2–8 kHz

• ϒ ratio: E1/ E5—where E1 is the energy in 0–1 kHz and E5 is the energy in 5–8 kHz.

The alpha ratio is the level difference between 0–1 and 1–5 kHz and indicates the spectral slope declination [13]. Hypofunctional voice production and soft phonation are characterized by a small alpha ratio, whereas relatively larger alpha ratios are noted in hyperfunctional voice disorders and loud phonation [13, 14]. ‘Resonant voice’is rich in harmonics, with a lower alpha ratio (less steep spectrum) [14].

Bhramari Pranayama, also known as humming breath, is produced with anterior/oral vibration while humming. The sound of Bhramari Pranayama is similar to those in vocal exercises that use resonant voices. Gently closing the ears with digits enhances the auditory feedback while producing Bhramari Pranayama. Resonant voice is proven to generate an 'optimum vocal economy' [15] and its applications in voice intervention programs are well documented [16–20]. Recently, integrating both traditional and modern health care techniques is gaining importance in health sciences. Traditional practices in India included Complementary and Alternative Medicine (CAM) such as herbal medicines, Ayurveda, Siddha, and Yoga, etc., for holistic well-being. Supporting evidence is emerging on the application of yoga and pranayama for general health [21–25] and vocal health [5, 6]. There is a need to establish a theoretical framework and descriptors for the CAM using available technology in voice and speech research. The findings may help practitioners working in voice and yoga to explore, describe and apply the pranayama techniques in vocal health care. The current study attempted to analyze the acoustic characteristics of Bhramari pranayama and estimate the closeness between Bhramari pranayama (traditional practice) and nasal consonant /m/ (i.e., produced similarly to the Bhramari pranayama; so labeled as modern practice). These comparisons between nasal consonant /m/ and Bhramari pranayama can take advantage in voice research and teaching (pedagogy) this type of pranayama.

Method

Design and Participants

The institutional ethics committee approved the study (Ref no: CSP/19/MAR/76/129). To avoid any gender influence on acoustic analysis measures, the investigator recruited only male subjects. Thirty-three adult male yoga trainers (Age range: 25–55 years; Mean: 39 years, SD: 9.00) who consented to participate were included in the study. Subjects recruited must be a certified adult male (< 55 years of age) yoga trainer from a private university or governmental organization with more than two years of experience teaching yoga. The subject must practice Bhramari pranayama on a regular day-to-day basis. Subjects with self-reported voice problems at the time of study/history of voice problems, h/o upper/lower respiratory tract (breathing disorders such as wheezing, COPD, pneumonia) infection, h/o psychological, cognitive, neurological, and hearing difficulties, and habits/history of smoking or consuming alcohol were excluded from the study. Subjects completed a subject information form on socio-demographic details, medical history, work experience, and practice duration of Bhramari pranayama.

Recording of Samples

Procedure

The subjects were instructed to sustain the nasal consonant /m/ and sound of Bhramari pranayama for 5–8 s in their comfortable voice. The investigator verified the utterance of Bhramari pranayama during the recording. Each token was repeated five times with a brief uniform period of silence (around 30 s) between the repeats. A total of (33 subjects × 5 (sound /m/) × 5 (Bhramari pranayama) repeats) 330 tokens were recorded and analyzed.

Recording Environment and Instrumentation

These samples were recorded at 20 kHz sampling frequency and 12-bit quantization using Computerised Speech Lab, Kay Pentax Model 4500b, and microphone (SM48) mounted on a stand positioned 15 cm away from the subject's mouth. 4–6 s sample was selected for analysis based on the perceptual judgment of a steady-state phonation with adequate loudness. The following acoustic parameters were analyzed:

1. Formant frequencies (fF1, fF2, fF3 & fF4)

2. Formant bandwidths (BF1, BF2, BF3 & BF4)

3. Antiformant

4. Alpha and beta ratio.

Sample Analysis

Estimation of Formant Frequencies and Bandwidth

The first four formant frequencies in Hz (fF1, fF2, fF3, and fF4) and formant bandwidth in dB (BF1, BF2, BF3, and BF4) values were obtained by using Linear Prediction Coding (LPC) analysis. The samples were displayed as a waveform in the source window. LPC based time history formant (fF1, fF2, fF3 & fF4) and bandwidth (BF1, BF2, BF3 & BF4) were generated for the selected portion of the sample. Figure 1 represented the first (fF1), second (fF2), third (fF3), and fourth peak (fF4) of the spectral curve. First bandwidth (BF1) was obtained by calculating the 3 dB difference from the formant peak, and consecutively BF2, BF3, and BF4 were also calculated.

Fig. 1.

Linear Prediction Curve (LPC) waterfall visualization of formants in CSL

Estimation of Anti-formants

Anti-formant (AF) in Hz is represented as a light band between the first dark band (fF1) and the next dark band (fF2). Figures 2 and 3 represented the spectrographic representation of the first and second anti-formants, respectively. A cursor was placed on the center-most point of the steady-state anti-formant band (first light band from the bottom), and the corresponding frequency displayed in the software was noted and entered into the tabulation sheet.

Fig. 2.

Spectrographic visualization of first antiformant for Bhramari pranayama in CSL

Fig. 3.

Spectrographic visualization of first antiformant for nasal consonant /m/ in CSL

Estimation of Spectral Energy

The samples were displayed as a waveform in the source window. FFT-based energy spectrum was used to measure the energy differences across frequencies. The energy in the frequency region 0–1 kHz was represented as E1, energy in the frequency region 1–5 kHz was represented as E2, energy in the frequency region between 0–2 kHz was represented as E3, and energy in frequency region between 2–8 kHz was represented as E4 respectively. The α ratio and β ratio were calculated using the formula given below:

$$\mathrm{\alpha }\text{ratio}=\text{E1}/\text{E2;}$$

$$\mathrm{\beta }\text{ratio}=\text{E3}/\text{E4}.$$

The ratios were calculated and entered into a spreadsheet.

Data Analysis and Statistics

The primary investigator analyzed the tokens for acoustic parameters. For reliability analysis, 10% of the samples were verified by the investigator (intrasubject reliability) and another lab associate (inter-subject reliability) at two different instances (one-week time gap between analyses). The intraclass Correlation Coefficient (ICC) test was used to assess the reliability. The intra- and inter-subject reliability values were 0.999 and 0.998, respectively, indicating good reliability. Shapiro–Wilk test of normality was done to check the distribution of the sample. As the normality was rejected, the values were converted to a logarithmic scale to standardize for the non-normality presented in the data. A generalized linear model and linear regression analysis were applied to account for random effects of subject variance.

Results

Within-subject variability for repeated trials was analyzed and did not show a statistically significant difference for formant frequencies, bandwidth, antiformant, and alpha & beta ratios. The data on within-group differences is listed as supplementary material. Table 1 showed the antilog mean and 95% confidence interval of first (fF1), second (fF2), third (fF3), and fourth (fF4) formant frequencies (in Hz) of nasal consonant /m/ and Bhramari pranayama. Antilog mean fF1 of both the nasal consonant and Bhramari pranayama was 242 Hz. Antilog means of fF2, fF3, and fF4 of nasal consonant /m/ was 1733, 2896, and 3815 Hz, respectively. On the other hand, the second, third and fourth formant frequencies of Bhramari pranayama were 1524, 2797, and 3739 Hz, respectively. There was a significant difference between both the groups for fF2 (p = 0.03). Figures 4 and 5 showed the highest, lowest and average antilog values of first (fF1), second (fF2), third (fF3), and fourth (fF4) formant frequencies of nasal consonant /m/ and Bhramari pranayama, respectively. The spacing between formants was around 1100–1200 Hz for nasal consonant /m/, and was around 1200 Hz for Bhramari pranayama.

Table 1.

Antilog mean (95% confidence interval) and p value of first (f_F1), second (f_F2), third (f_F3) and fourth (f_F4) formant frequencies (in Hz) for nasal consonant /m/ and Bhramari pranayama

Formant frequencies	Nasal consonant /m/		Bhramari pranayama		p-value
	Antilog mean	95% CI	Antilog mean	95% CI
fF1	241.8	228.7, 255.0	241.6	229.5, 253.7	0.97
fF2	1733	1591.8, 1874.2	1523.8	1386.7, 1660.9	0.03*
fF3	2896.3	27,573, 035.70	2796.9	2676.3, 2917.5	0.27
fF4	3815.4	3701.3, 3929.4	3738.8	3612.9, 3864.7	0.36

*Significant at p < 0.05

Fig. 4.

Highest, lowest and average antilog values of first (f_F1), second (f_F2), third (f_F3) and fourth (f_F4) formant frequencies of nasal consonant /m/ obtained from male yoga trainers

Fig. 5.

Highest, lowest and average antilog values of first (f_F1), second (f_F2), third (f_F3) and fourth (f_F4) formant frequencies of Bhramari pranayama obtained from male yoga trainers

Table 2 showed the antilog mean and 95% confidence interval of first (BF1), second (BF2), third (BF3), and fourth (BF4) bandwidth (in Hz) of nasal consonant /m/ and Bhramari pranayama. Figures 6 and 7 showed the range, i.e., highest, lowest, and average values of first (BF1), second (BF2), third (BF3), and fourth (BF4) bandwidth of nasal consonant /m/ and Bhramari pranayama, respectively. Antilog mean of the BF1 for both nasal consonant /m/ and Bhramari pranayama was 165 Hz, BF2 was around 262 Hz and 305 Hz, BF3 was observed at 226 Hz and 270 Hz, and BF4 was 239 and 270 Hz. There was a significant difference between both the groups for BF3 (p = 0.02). There was no significant difference in BF1, BF2, and BF4 between the two groups.

Table 2.

Antilog mean (95% confidence interval) and p value of first (B_F1), second (B_F2), third (B_F3) and fourth (B_F4) bandwidth (in Hz) for nasal consonant /m/ and Bhramari pranayama

Bandwidth	Nasal consonant /m/		Bhramari pranayama		p value
	Antilog mean	95% CI	Antilog mean	95% CI
BF1	165.1	150.9, 179.3	165.4	145.4, 185.5	0.97
BF2	262.8	237.6, 288.0	305.6	268.6, 342.7	0.05
BF3	226.3	206.3, 246.2	270.4	237.7, 303.2	0.02*
BF4	239.2	217, 261.3	270.4	237.7, 303.2	0.15

*Significant at p < 0.05

Fig. 6.

Highest, lowest and average antilog values of first (B_F1), second (B_F2), third (B_F3) and fourth (B_F4) bandwidth of nasal consonant /m/

Fig. 7.

Highest, lowest and average antilog values of first (B_F1), second (B_F2), third (B_F3) and fourth (B_F4) bandwidth of Bhramari pranayama

Table 3 showed the antilog mean and 95% confidence interval of antiformant (in Hz) of nasal consonant /m/ and Bhramari pranayama. The antilog mean values of nasal consonant were 750 Hz and of Bhramari pranayama was 705 Hz. The values were statistically lower in Bhramari pranayama compared to nasal consonants (p = 0.01). Table 4 showed the antilog mean and 95% confidence interval for alpha and beta ratio. The alpha ratio was low, and the beta ratio was high in Bhramari pranayama compared to the nasal consonant. There was no statistically significant difference in alpha and beta ratio between nasal consonant /m/ and Bhramari pranayama.

Table 3.

Antilog mean (95% confidence interval) and p value of antiformant (in Hz) for nasal consonant /m/ and Bhramari pranayama

Antiformant frequencies	Nasal consonant /m/		Bhramari pranayama		p value
	Antilog mean	95% CI	Antilog mean	95% CI
Antiformant	750.3	727.1, 773.6	705.3	682.7, 727.9	0.01*

*Significant at p < 0.05

Table 4.

Antilog mean (95% confidence interval) and p value of alpha and beta ratio (in Hz) for nasal consonant /m/ and Bhramari pranayama

Alpha and beta ratio	Nasal consonant /m/		Bhramari pranayama		p value
	Antilog mean	95% CI	Antilog mean	95% CI
Alpha ratio	21.5	(− 7.4, 50.4)	9.7	5.9, 13.6	0.39
Beta ratio	8.6	5.0, 12.1	11.1	1, 21.2	0.60

Discussion

Formant Frequency

Results of the study revealed that there was a significant difference between nasal consonant /m/ and Bhramari pranayama in fF2, BF3, and antiformants; thus rejecting the null hypothesis. In the current study, nasal consonant /m/ and Bhramari pranayama had lower first formant frequencies at about 241 Hz. The finding from the study was similar to that of Maeda [26]. The nasal cavity picks up the vibration and contributes to nasal resonance producing a 'nasal murmur'. The nasal murmur generally has a low formant frequency around 200–300 Hz. Low frequency of the nasal sounds is contributed by the resonance of large nasal passages constricted by narrow nasal opening [27]. The volume and impedance of the sinus cavity also played important roles in contributing to the lowest formant frequency in nasal sounds [28].

Second formant frequencies (fF2) for nasal consonant /m/ were within the range of frequencies mentioned by Chen [29] and Fujimura [27]. fF2 depends on the size of the narrow velar passage, thereby producing a lower formant frequency [9]. Bhramari pranayama had a lower fF2 mean value compared to the nasal consonant. Lower fF2 in Bhramari pranayama could be attributed to a narrower velar passage in contrast to nasal consonant /m/. A decrease in the cross-sectional area of the vocal tract would increase air column inertance and facilitate resonant voice [30]. fF1 relates to the pharyngeal cavity volume, and fF2 relates to the front cavity volume (oral articulators) [10]. Physiologically, it can be assumed that Bhramari pranayama has a larger pharyngeal cavity, i.e., lower fF1 when compared with nasal consonant /m/.

The third formant frequency range (2006 Hz–4311 Hz) of the nasal consonant in this study is similar to the range specified by Fant [31] and Fujimura [27]. The resonance of the third formant frequency seems to be chiefly dependent on the characteristics of the pharynx cavity [10, 32].

From Fig. 4, the nasal spacing of formants for /m/ was around 1100- 1200 Hz between formants in this study, which is in agreement with Tabain et al. [33]. In this study, formant frequency values of nasal consonants were within the range previously reported by Fujimura [27]. A strong low-frequency nasal formant is associated with the vocal tract extending up to the nasal cavity for men [27]. Mean values of fF1, fF2, fF3, and fF4 of nasal consonant /m/ are comparatively higher than Bhramari pranayama; this may be attributed to the size of constriction and placement of articulators. The location of formant frequencies considered above was possibly also dependent on individual speakers.

Bandwidth

From Figs. 6 and 7, it can be visualized that Bhramari pranayama had overall higher mean values of bandwidths than nasal consonant /m/ except the first bandwidth. Several factors influence bandwidths. These are effects of radiation of sound in the oral/ nasal cavity, compliance of vocal tract walls, viscosity, heat conduction, and glottal opening [14, 31]. These factors may also vary according to the speaker characteristics such as age, gender, health status, and specific control of strain and relaxation of vocal fold musculature during phonation [14]. The oral cavity for /m/ has an abrupt termination, and the oral cavity has a smaller ratio of surface area to volume [32]. The Bhramari pranayama may have considerably more damping than that of /m/ because of the way it is produced (softer intensity level and with relaxed pharyngeal/ oral cavities), thus resulting in increased acoustic energy loss at higher frequencies. This is reflected as higher bandwidths in BF2, BF3, and BF4 compared to that of nasal consonant /m/. The lining of nasal passages may also contribute to absorbing a considerable amount of acoustic energy. Increased damping may contribute to wider bandwidths. High bandwidth values echo Tabain et al. [33] findings that showed palatal sounds had higher bandwidth (BF2- BF4) compared to bilabial, velar, and alveolar. This may be related to the large coupling area at the velar entrance, as suggested by Fant [34] for palatalized sounds. This may also mean that Bhramari pranayama is relatively more palatalized, i.e., tongue position may be higher than that in /m/, simultaneously increasing the pharyngeal cavity dimension. Further, researchers can validate these findings with imaging studies to understand the positioning of the articulators.

Antiformants

As higher formants were not discernible clear in the spectrograms, the presentation was restricted only to the first antiformant. Bhramari pranayama had a lower mean value of antiformant compared to nasal consonant /m/, which was statistically significant. Antiformant of nasal consonant /m/ was in the range between 571 and 1008 Hz as proposed by Fujimura [27]. The frequency of antiresonance relates to the size of the oral pharyngeal cavity i.e., larger cavity yields lower frequencies and smaller cavities, higher frequencies [10]. The lower antiresonance frequency of Bhramari pranayama probably indicated a larger pharyngeal-oral cavity when compared to that in the production of the nasal consonant /m/.

Alpha and Beta Energy Ratio

Nasal consonant /m/ had a lower alpha ratio and higher beta ratios compared to Bhramari pranayama. A higher beta ratio of nasal consonant /m/ can imply the presence of rich higher harmonics that of Bhramari pranayama's sound. A lower beta ratio in Bhramari pranayama may be attributed to the reduced intensity during the production of sound, thus reflecting very low energy in higher harmonics. A resonant voice rich in harmonics has a less steep spectrum with fewer alpha ratios [30]. Bhramari pranayama revealed a low alpha ratio (similar to nasal consonant /m/) and a very low beta ratio (0.5—very low when compared to nasal consonant /m/—approximately 4.6). Verdolini [35] mentioned that a resonant voice is produced as a clear voice quality with less effort. This may be somewhat similar to the way Bhramari pranayama is produced. Verification at the level of the glottis and with aerodynamic parameters may ascertain the above proposition.

Summary and Conclusion

The present study described the acoustic characteristics of Bhramari pranayama and documented the differences between nasal consonant /m/ and the sound of Bhramari pranayama. The results of the study imply a significant difference in second formant frequency, third formant bandwidth, and antiformant, indicating that the size of constriction and placement of articulators could be different from that of nasal consonant /m/. Change in size and constriction could be due to the relaxed oral/ pharyngeal articulator that was enhanced by the auditory feedback while producing Bhramari pranayama. This proposition, however, needs more verification through other metrics. Future studies can focus on investigating Bhramari pranayama's application in different voice disorders, and airway/laryngeal disorders.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

No funding was received for conducting this study.

Declarations

Conflict of Interest

The authors declare that they have no conflict of interest.

Ethics Approval

The study was approved by the Institutional Ethics Committee, Sri Ramachandra Institute of Higher Education & Research (DU) (Ref no: CSP/19/MAR/76/129).

Informed Consent

Informed consent was obtained from all subjects prior to the participation of the study.