Exploring Breathing-Music Coupling: Using the Breathing Mirror for Somatic Reflection in Piano Performance

AI summary

Overview

Research area: Human-Computer Interaction / New Interfaces for Musical Expression (NIME), specifically wearable physiological sensing and soma design applied to piano performance.

Technical level: Intermediate. The conceptual framing is accessible, but the paper includes specific signal-processing parameters (filter orders, thresholds, sampling rates) and qualitative research methodology that assume some familiarity with HCI or music technology research.

Scope: A single-participant, four-week longitudinal study introducing the "Breathing Mirror," a post-hoc visualization interface that synchronizes respiratory signals from wearable strain-sensor belts with MIDI, audio, and annotation data to support somatic reflection in piano performance.

What This Paper Is About

For pianists, breathing is an automatic and largely invisible process — unlike wind players or singers, they do not need breath to produce tone, so respiratory patterns stay below conscious awareness and are hard to recall or refine. Sensors can record precise breathing data, but they capture physical metrics rather than the performer's felt experience, and the high cognitive load of playing makes it nearly impossible for musicians to remember their internal states with temporal accuracy. This paper presents a system and methodology for externalizing that hidden somatic experience, letting a pianist review her own breath alongside her music.

Key Contributions

1. The Breathing Mirror system and methodology. An interactive interface integrating a Baseline View (synchronized physiological and musical signals), a First-Person View (subjective recall), and an Interpersonal View (collaborative researcher–performer reflection) to structure somatic reflection.

2. A breathing segmentation pipeline that extracts inspiration and expiration phases from two-channel wearable sensor data using filtering, adaptive peak detection, and multi-channel fusion, validated by manual researcher refinement and performer verification.

3. Four somatic themes proposed regarding the generalizability of breathing-music coupling, intended as a foundation for future evaluation with a broader cohort of expert pianists.

4. An empirical demonstration through a four-week longitudinal study with one skilled amateur pianist (35 years of experience), showing that the system reveals patterns of breathing-music coupling and identifies blind spots where objective data diverges from subjective perception.

Main Findings

• Six breathing patterns were identified in the performance. Thematic analysis produced six primary themes characterizing the breathing-music relationship, ranked by frequency: Breath-holding, Sectional Transitions (preparation for changes in musical scenes), Expression Marks (responses to specific score annotations), Beginning of Major Musical Structures, Foot Pedaling Actions, and Preparing for Bass Notes. Exact frequency counts are not reported in the text; Figure 4 visualizes their distribution across the performance duration with circle diameters scaled to breathing-cycle amplitude.

• Note on theme count: The abstract and contribution list refer to "four somatic themes," while the findings section reports "six primary themes." The paper does not reconcile this discrepancy in the provided content.

• Inhalation often prepares left-hand bass notes. YK described this as a core habit: "it's mostly matching the [left hand] bass as the main pattern." In arpeggiated left-hand passages, the inhalation aligns with the harmonic foundation — "for the lowest note of an arpeggio, the inhalation prepares for the bass note of the left hand."

• Breathing marks structural and expressive transitions. YK inhales deeply at major section entries to establish timing, and increases inhalation depth when a phrase change coincides with a bass note. She reported needing "a deep breath" when "entering a new scene" because "the mood of the song changes."

• Breath-holding appears in expressive and physically demanding passages. During ritardando, YK adopts a controlled, suppressed breathing style ("killing my breath to play carefully"). She also attributes breath-holding partly to a habit from wind instrument practice.

• Physiological constraints force respiratory substitution. When a phrase is too long or notes too close, YK substitutes exhalation for the intended inhalation: "I want to inhale, but the next note is near; instead of inhaling, I exhale." On long melodies, she said matching breathing to the phrase would mean she would "die [of breathlessness]."

• Awareness diverges from measured reality. Some inhalations were more pronounced than YK expected, functioning as a physiological "reset" ("now I'm going to play"). Micro-inhalations sustained melodic continuity across multi-measure phrases.

• Timing of inhalation shifted ahead of musical events. YK noted her inhalation occurred slightly ahead of the corresponding musical event, particularly when approaching high-register passages — she felt she was inhaling "at the beginning" but the data showed it was "a bit more ahead."

• A proposed parallel between breathing and pedaling. YK suggested her inhalation/exhalation cycle mirrors pedal pressing and releasing to manage phrase clarity.

• A concrete annotated example: YK identified a "big inhale" triggered by a chord change from B to C at the 38.7-second mark, logged as "change of chord, big inhale B-C."

• Small but regular gains from the data-mining setup. — Not applicable; this paper does not involve data mining.

Methodology in Plain English

The researchers built a wearable breathing measurement setup using two textile belts containing carbon-nanotube strain sensors, positioned at the mid-thorax (lower rib cage) and the umbilical level (abdomen) to monitor thoracic and abdominal displacement separately. The belts were connected to a data acquisition unit sampling at 10 kHz, while a Yamaha Stage Piano CP88 captured MIDI at roughly 1 ms resolution and audio at 44.1 kHz. All streams were synchronized with a common clock (Rosendahl mif4) and transferred to a 13-inch MacBook Pro for visualization.

The software pipeline cleaned the raw breathing signal with a fifth-order Butterworth band-pass filter (0.05–2.0 Hz), applied zero-phase forward-backward filtering to keep respiratory onsets aligned with musical events, and smoothed the result with a moving average window of 15. Inhalation and exhalation events were detected using an adaptive prominence threshold set to max(0.05, 0.2 × standard deviation of the smoothed signal). Because pilot testing showed the thoracic channel had a higher signal-to-noise ratio, thoracic onsets were treated as primary; an event was validated either if a corresponding abdominal peak appeared within 0.75 s, or if its relative amplitude exceeded a confidence threshold of 0.7 standard deviations. Brief "trough-peak-trough" patterns shorter than 0.6 s were discarded as movement noise. Critically, the algorithm was used only as an assisted-segmentation tool — all detected events were manually refined by researchers and verified by the pianist.

For the study itself, the team ran four weekly sessions with one pianist, YK, a skilled amateur with 35 years of classical training (20 years of classical training with little emphasis on breathing, then 15 years of intensive breath-centered practice). Session one (30 minutes, online) covered familiarization. Session two (120 minutes) split into 60 minutes of performance recording (warm-up, practice runs, 2–3 recorded takes) and 60 minutes of overview annotation, with YK seated at the piano touching the relevant phrases while watching the interface. Sessions three and four (60 minutes each) involved in-depth micro-annotation of the full piece, with YK engaging the interface away from the piano. The piece was the opening section of Nausicaä of the Valley of the Wind by Joe Hisaishi — a three-minute work with six sections alternating dense rhythmic ostinatos and expansive lyrical melodies.

Analysis used inductive thematic analysis following Braun and Clarke's reflexive framework. Interviews were conducted primarily in Japanese, transcribed with Microsoft Word's transcription service, translated to English via DeepL, and jointly audited against the original audio by both authors. Annotated timestamps were compiled into a unified CSV file to triangulate observations, calculate the frequency of breathing-music coupling behaviors, and generate a timeline visualization.

Why This Matters

Impact on research: The work reframes breathing from an internal biological function into an articulate expressive parameter, and offers a replicable three-lens methodology (Baseline, First-Person, Interpersonal) for studying physiological signals in performance. It directly addresses the "kinesthetic mismatch" — the documented gap between the brain's internal representation of an action and the body's actual physiological state — and demonstrates a way to surface that mismatch for performers and researchers alike. Most prior breathing-aware NIME work has focused on vocal or wind contexts where respiration is mechanically required for sound; this extends the paradigm to an instrument where breathing is not needed for tone production.

Real-world applications:

• Piano pedagogy. Teachers could use synchronized breathing-music visualization to make respiratory habits visible to students, particularly habits students cannot consciously report.

• Music practice and self-study tools. A post-hoc reflection interface lets performers compare what they felt during a performance with what was physiologically recorded, supporting deliberate practice on expression rather than just accuracy.

• Wearable sensing in performance contexts. The textile-integrated CNT strain sensor's linearity, thinness, and low spring constant make it suitable for embedding in garments without disturbing a musician's gestures.

• Clinical and rehabilitation use. The paper notes the same breathing belts have already demonstrated effectiveness in medical contexts, extending the technology beyond music.

Industry relevance: The sensor is GummiStra™ from Yamaha Corporation, and two co-authors are affiliated with Yamaha — indicating direct commercial interest in wearable physiological sensing for music products and pedagogy tools. The interface also points toward a future category of practice software that monitors internal bodily states rather than only external correctness (pitch accuracy, rhythmic timing), which the paper notes traditional piano learning interfaces emphasize.

Future Directions

• Large-scale validation with a broader cohort. The authors explicitly propose the four somatic themes as a foundation for evaluation "across a broader range of pianists" and "a broader cohort of expert pianists." The current idiographic single-participant design intentionally prioritizes depth over statistical sampling.

• Optimizing the automated segmentation model. The authors state that future iterations "will continue optimizing the automated model," which currently serves as an assisted-segmentation tool requiring full manual researcher refinement and performer verification.

• Extended reflection content. The provided text is truncated mid-sentence in Section 5.4, and a further findings section appears to follow; questions raised by the completed 5.4 discussion are not fully reported here.

• Open question — reconciling objective data with somatic authority. The study documents several divergences between what YK felt and what the sensors recorded (temporal shifts ahead of musical events, unexpected amplitude). How such divergences should be interpreted pedagogically or analytically remains an open issue the paper does not resolve.

Target Audience

• HCI and NIME researchers working on wearable sensing, soma design, or embodied interaction.
• Music performance researchers studying breathing, expression, and physiological correlates of musical structure.
• Piano pedagogues and advanced piano teachers interested in making respiratory habits and expressive intent visible to students.
• Musicians with high somatic awareness, particularly those who have undergone breath-centered training and want tools for structured self-reflection.
• Designers of music practice technology looking beyond pitch and timing feedback toward internal physiological states.

Readers seeking statistically generalizable results or multi-participant validation should note that this is a single-participant, idiographic study, and the paper positions its themes as hypotheses for future testing rather than established findings.

Authors’ abstract

While breathing is essential to living and for sound production in some instruments, for pianists, it is often a hidden and automatic process, making it difficult to analyze or refine. A critical gap exists between data and awareness: while sensors record precise physical metrics, they fail to capture the performer's somatic experience. Conversely, the high cognitive load of performance makes it nearly impossible for musicians to recall their internal states with temporal precision. To address this, we present a system, Breathing Mirror, and associated methodology designed to externalize the pianist's internal somatic experience through three analytical lenses: a Baseline View (synchronized signals), a First-Person View (subjective recall), and an Interpersonal View (collaborative reflection). Through a four-week longitudinal study with a skilled amateur pianist (35 years of experience), we evaluated the system's effectiveness by recording respiratory data using textile-integrated strain sensor belts. The results show that the Breathing Mirror reveals some patterns of breathing-music coupling and identifies critical blind spots where objective data diverges from subjective perception. Furthermore, we propose four somatic themes regarding the link between breathing and musical elements, offering a foundation for future large-scale validation across a broader range of pianists. This work provides a new way to study body signals, transforming breathing from an internal biological function into an articulate expressive parameter.