HAMR 2013 Proceedings

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Respiration Phase Coordination ======

| Authors | Finn Upham |
| Affiliation | New York University |
| Code | aha... |

If music can encourage stimulus synchronous respiration, what part of the respiration cycle is being aligned to the music? 

Using respiration band stretch data from one listener recorded during multiple listenings (24) to the same set of musical stimuli (25), significant coordination is measurable on inspiration onsets for some pieces. Is any other part of the respiration cycle better coordinated?

===== Part 1: approximating the phase cycle of respiration =====

The figure below shows three representations of respiration from the subject during the presentation of a single piece of music. First is the filtered stretch sensor data reported over time; each breath takes a few seconds, inspiration during the rise in value, exhilation during the fall, and period of relative stability of variable length in bet ween. Second is the onsets of inspiration, as selected by my less than perfect detection algorithm which selects local minima, calculates the duration of time between them, and excludes those which are too close together. And last is the approximate phase values, mapping the respiration cycles to the interval [0,1] by fitting the model cycle to the sensor data between inhilation onsets.

{{:exampleresp.png?800|Three representations of respiration over time (seconds) from one subject during the presentation of a musical stimulus.}}

To define the phase cycle, six characteristic points were chosen as anchors: 
  * 0 : Inspiration onset
  * 0.175 : Max inspiration rate
  * 0.4 : Peak volume
  * 0.6 : Max exhilation rate
  * 0.9 : Stablised volume (before next inspiration)
  * 1 : End of respiration cycle (coinciding with next inspiration)

The figure below shows one typical respiration cycle (blue) with these anchors marked with asterixes. The other lines are the associated functions by which these points were selected: the first and second derivatives, and a special combination of the absolute value of these two for identifying the stable tail.

{{:phasedef.png?500| Example of the phase anchors for one respiration cycle. Blue is the respiration stretch gauge data, red first derivative (rescaled), yellow second derivative (rescaled), pink a combination of the two. }}

The phase values for these anchors were selected empirically by rounding the median values of anchor times relative to cycle duration for each across all respiration cycles in the data set (~34 hr of breathing while listening). 

===== Part 2: Coordination of different phases =====

While any phase value could be tested for coordination, I chose the first five anchor points to roughly check which were the most stimulus-synchronous component of the respiration cycle. For any given stimulus, the coincidence of a given phase component across listenings was estimated at each time point using Finn's special mix of activity analysis and Unitary Event Analysis. This process estimates the local likelihood of the number of sessions reporting that specific respiration phase in a half second time frame. The phase components are compared by calculating the ratio of time frames with high coincidence (likelyhood < 0.025) and very low coincidence (likelyhood > 0.975). 

Below are the results for two pieces which exemplify two differing patterns in the across the stimuli.

==== Romance (Harp and Bassoon chamber work by C. Debussy) ====
Like some of the other pieces which showed strikingly high coordination in the earlier analysis of inspiration onsets, this early phase component continued to be the most coordinated of the cycle. These pieces feature prominent vocal lines or, in this case, a wind instrument, in which the listener likely interpreted the performers respiration cycle. More on this issue can be found [[http://finnupham.wordpress.com/2013/05/13/respiration-coordination-take-n1/|here]].

{{:romanc.png?800|}}

This figure shows first in rows the respiration stretch data over time for all 24 of this subject's listenings to this piece of music, with yellow showing inhilations, dark green exhilations, and light green no change in stretch. Second graph shows the ratio of sessions reporting a given phase in each half second time frame; predictably, the phase components vary closely together over time with a short offset between them. Third is the local surprise values for these coincidences in phase, surprise being a function of the likelihood in which big positive values means remarkably high numbers of sessions in that specific respiration phase and big negative values represent remarkably few sessions in that phase. And last is a histogram of the surprise values across time for each phase value. 

Phases
    __0__ 0.175 0.4 0.6 0.90

High and Low Coincidence rates (>0.025)
    __0.1110__    0.0932    0.0899    0.0783    0.0870
    __0.0610__    0.0476    0.0370    0.0408    0.0303
Total extreme coordination
    __0.1720__    0.1408    0.1269    0.1192    0.1173
    
With these numbers we see more coordination than would be expected by chance for all phase components, but the strongest coincidences occur at the onset of the respiration cycles, suggesting that the music is coordinating this component of the listener's breathing. 

==== Gong Hotel (Acadian Hip Hop by Radio Radio) ====
But in contrast to the above, some pieces previously interpreted as having little respiration coordination showed higher coincidence rates in the exhilation part of the cycle. This might be the result of the participant exhaling rhythmically to highly danceable pieces.

{{:gonghotel.png?800|}}

The figure follows the same format as the one directly preceeding, and some differences in pattern are worth noticing. In the top graph, the respiration band change data shows intervals of fast and small oscillations, in contrast to the slow breaths of the previous example. This is related to the participant's motor engagement with the music, i.e. dancing in her chair. 

Phases
    0 0.175 0.4 __0.6__ 0.90

High and Low Coincidence rates (>0.025)
    0.0795    0.0835    0.0819    __0.0919__    0.0722
    0.0463    0.0445    0.0454    __0.0501__    0.0332
Total extreme coordination
    0.1257    0.1280    0.1273    __0.1420__    0.1054

Though not as many frames have extreme coordination as the last example, the alignment of the point of max exhalation is notably higher than the other phases, in both low and high coincidence. It would be useful to look at how these exhalations align with the beat of the music: is this a case of percussive breathing, or some more subtle compromise between the listener's respiration and the music?


===== Conclusion =====

Respiration is a complex phenomenon, satisfying involuntarily controlled metabolic processes and quasi-voluntary expressive behaviours simultaneously. Within the later, stimulus synchronous respiration may occur as the results of multiple competing bodily expressions of engaged music listening. Further work should will systematically evaluate all the stimuli in the data set and repeat these investigations in data collected from other future subjects for confirmation.
HAMR 2013 Proceedings

User Tools

Site Tools

Page Tools
