Penelope Bekiari
Featured Projects
SoundCloud:
Hyponoia is an AI-assisted neuroadaptive music system developed as part of my PhD research. It combines real-time EEG analysis, Max/MSP, Python, OSC communication, and adaptive generative music within a human–AI interaction framework. The prototype analyses the performer's neural activity and generates evolving soundscapes through computational musical decision-making. It also incorporates human feedback to gradually adapt future generations. Potential applications include intelligent musical interfaces, computational creativity, adaptive performance systems, wearable sensing, accessible music technology, and human–computer interaction. Future work explores neuroadaptive interaction for people with visual, motor, and communication impairments. Current prototype: uses a small, curated sample library for demonstration purposes.
1. This video provides a concise overview of recent creative and research projects.The opening section presents Rhoē, a performance for piano or cello, and live electronics that incorporates the Hyponoia neuro-responsive biofeedback system. During live performance, the system analyses the performer's EEG activity in real time, classifies cognitive states, and communicates with Max/MSP to influence the evolving musical environment. As part of my doctoral research, the system is currently being extended into an AI-assisted generative soundscape engine capable of adaptive musical behaviour. The performance is accompanied by visual projections that explain to the audience when and how neurophysiological states change throughout the performance.
2. The second section demonstrates a wearable gyroscopic sensing system that allows interaction with sound through spatial movement. In the demonstration, hand movements are mapped to different regions of a painting by Claude Monet, with the resulting spatial orientation controlling the soundscape in real time. Although shown as a hand-worn prototype, the system has been designed primarily as a head-mounted wearable interface intended to support visually impaired users in exploring paintings through spatial audio and interactive sonification.
for trumpet, electronics, and bio-electronics
Permeable Signal explores the boundaries between the human body, the instrument, and the technological environment through a live performance. The trumpet acts as a sonic core, with breath, articulation, and its metallic resonance serving as the primary material. This material passes through a network of electronic processing and bio-signals (EEG, heart rate, electrodermal activity), which erode, amplify, or transform the sound.
The notion of permeability forms the axis of the composition: the natural and the artificial do not stand in opposition but infiltrate one another, exchanging energy and continuously transforming. In this way, an organic, abstract form emerges in real time: the sound of the trumpet becomes an electronic signal; the signals of the body become music; and the boundaries between the organic and the mechanical remain fluid. Video here!
Wavelets / Rustle for cello, electronics, brainwave and heartbeat data, field recordings, and live processing. Wavelets and Rustle are two interconnected works that explore the relationship between human perception, environmental sound, and musical expression. In both pieces, real-time brainwave and heartbeat data influence the performance by controlling the timbral colour of the cello and determining the structure and character of the electronic soundscape that surrounds it. Through this system, the performer’s internal physiological states directly shape the evolving sonic environment, creating a feedback loop between mind, body, and sound.
Rhoē explores the fluid interplay between sound and the human body through Hyponoia, an AI-assisted neuroadaptive music system developed as part of my PhD research. Written for piano and live electronics, the work integrates physiological data—including EEG activity, heart rate, and galvanic skin response—into a real-time computational performance environment.
Within Hyponoia, biometric signals are analysed continuously and interpreted into higher-level computational states that shape adaptive electroacoustic processes. Rather than directly controlling isolated sound parameters, the system influences generative musical behaviour, allowing the electronic soundscape to evolve alongside the performer's physiological state.
The piano's acoustic gestures merge with dynamically generated electronic textures, creating an evolving dialogue between performer, physiology, and artificial intelligence. Through neuroadaptive interaction and AI-assisted sound generation, Rhoē investigates new forms of embodied musical expression in which biological processes become an active component of composition and performance.
HYPONOIA
(How the Works Above Were Created)
Embodied Co-Creation, Generative Memory & Adaptive Learning in Biointeractive Musical Performance
Hyponoia is an evolving biointeractive musical research system that explores how physiological activity, listening, sound, and adaptive computation can participate in a shared compositional process.
Developed through ongoing artistic research in electroacoustic composition, physiological computing, and augmented performance, Hyponoia integrates real-time EEG analysis, Max/MSP, Python-based generative processes, curated sonic memory, and adaptive learning.
Rather than treating the body as a controller or translating individual brain signals directly into isolated musical parameters, the system interprets physiological variation through a layer of symbolic performance states. These states modify broader musical behaviours and influence how sound is selected, transformed, organised, and returned to the performer.
Hyponoia does not attempt to decode thoughts, identify emotions, or claim direct access to subjective experience. Physiological signals are treated as dynamic compositional material: changing conditions that can influence musical form without completely determining it.
From Biofeedback to Embodied Co-Creation
The first architecture of Hyponoia emerged from research into neuro-responsive performance and biofeedback.
EEG activity is analysed in short temporal windows and interpreted relative to an adaptive physiological baseline. Activity across theta, alpha, beta, and gamma ranges contributes to a set of symbolic musical states, including neutral, alpha–theta, beta–gamma, beta–alpha, alpha–gamma, and theta-dominant states.
Instead of producing simple one-to-one mappings, these states activate families of musical behaviour such as sustained resonance, granulation, recursion, reverse delay, phrase looping, and temporal stretching.
The important distinction is that the performer does not simply “control” the electronics.
Physiological activity changes the conditions of the musical environment. The resulting sound returns to the performer through listening, potentially affecting attention, gesture, interpretation, and subsequent physiological activity. Performance therefore develops as a recursive relationship between body → system → sound → listening → body.
The current research extends this principle beyond responsive processing toward adaptive generative co-creation.
An Inspectable Human–AI Architecture
Hyponoia combines a Max/MSP live-performance environment with a set of Python processes responsible for physiological-state communication, sound-memory analysis, generative composition, computational evaluation, and persistent learning.
Communication between Max/MSP and Python takes place through Open Sound Control (OSC).
During performance, Max/MSP can communicate the current musical and harmonic situation to the generative engine and request a new soundscape. The Python system then constructs new material under those conditions and returns the generated audio to the Max/MSP performance environment.
The relationship is therefore bidirectional:
physiological activity → symbolic state → Max/MSP → generative engine → new soundscape → performer
The system is intentionally designed to remain inspectable. Rather than concealing musical decisions inside an opaque model, the principal stages of interpretation, generation, evaluation, and adaptation can be examined individually.
Generative Memory
The Hyponoia generative engine is not based on text-to-music prompting and is not trained on a large external music corpus.
Instead, it composes from a curated sonic memory created specifically for the project.
Recorded materials are segmented into individual sound objects and analysed according to acoustic and musical characteristics including energy, brightness, noise, attack, duration, pitch, harmonicity, resonance, musicality, richness, ambient potential, novelty, and transient behaviour.
Sound objects can then assume different provisional musical functions such as:
gesture — texture — resonance — noise — impact
These categories describe what a sound may do compositionally rather than what the sound represents.
Generation takes place through constrained probabilistic selection. The system considers the current musical state, harmonic compatibility, previous selections, continuity, learned preferences, exploration history, and the potential compositional role of each sound object.
The performer therefore does not specify every event, transformation, or formal decision. At the same time, the generator does not operate independently of the performer.
The result emerges from the interaction between the performer, the current physiological condition, the designed sonic memory, previous learning, and probabilistic computational decisions.
Adaptive Learning: Two Forms of Listening
A central part of the current Hyponoia architecture is the relationship between computational evaluation and embodied human listening.
After generating a soundscape, an internal computational critic analyses musical characteristics including:
musicality, coherence, richness, transitions, spectral bloom, and overall behaviour.
This critic does not “hear” music and is not treated as an autonomous aesthetic authority. Its measurements are heuristic descriptions of acoustic behaviour.
The composer then listens to the same generated material and evaluates the musical result.
Crucially, disagreement between computational evaluation and human listening is not treated as an error to be eliminated. It becomes part of the system's learning history.
If the computational system considers a passage highly coherent but embodied listening suggests otherwise, this difference can gradually modify the conditions used during subsequent generation.
Hyponoia therefore develops two interconnected forms of memory:
Global learning changes how the system evaluates qualities such as musicality, richness, coherence, transitions, spectral development, noise, or impact.
Sample-level learning changes the accumulated history and future probability of individual sound objects that have participated in previous renders.
Past encounters consequently leave traces without becoming absolute instructions.
Learning Without Complete Optimisation
Persistent learning introduces an important compositional problem.
If a system continuously reinforces materials that previously received positive evaluations, it may gradually become increasingly repetitive. A system that becomes very good at predicting past preference can also become less capable of producing something genuinely unexpected.
Hyponoia therefore places learning and exploration in deliberate tension.
The system remembers, but it also maintains mechanisms that allow less familiar material to re-enter the compositional field.
This is fundamental to the project's conception of co-creation.
The objective is not to train a machine to reproduce an idealised version of the composer's existing taste. Instead, the system should retain enough memory for previous interactions to matter while remaining sufficiently open for unfamiliar combinations, transformations, and sonic relationships to emerge.
Co-creation is therefore understood as a process of recursive alteration: the body changes the computational situation; computation reorganises musical memory; sound changes the conditions of listening; human evaluation leaves traces within the system; and exploration prevents those traces from becoming fixed rules.
The human and computational participants are neither identical nor equally intentional.
Musical agency emerges temporarily across bodies, algorithms, memory, sensing technologies, sound materials, and acts of listening.
Hyponoia in Performance
Hyponoia has developed through practical experimentation in electroacoustic and biointeractive performance.
Rhoē, for piano and adaptive live electronics, is one performance context in which physiological information—including EEG activity—participates in shaping the behaviour of the electroacoustic environment.
The generative research is now extending this performance architecture toward soundscapes that can be constructed dynamically from Hyponoia's evolving sonic memory, rather than relying exclusively on predetermined processing structures.
The project remains an evolving artistic-research prototype, with development continuing across neuro-responsive performance, adaptive generative composition, embodied human–AI interaction, and persistent musical learning.
Listen — Generative Training Progression
Living Ecosystem — D5 | Baseline Training
An early representative render from the adaptive generative sound engine.
Living Ecosystem — D5 | Intermediate Training
A later stage of adaptive training following iterative human evaluation.
Living Ecosystem — D5+ | Extended Training
An extended training state demonstrating the evolving relationship between sonic memory, exploration, and adaptive evaluation.
Code & Open Research
Hyponoia — Max/MSP
Biointeractive live-performance environment and system integration.
https://github.com/penbek7-maker/hyponoia-maxmsp
Muse Microstates
EEG processing and neuro-responsive state-analysis tools associated with the Hyponoia research framework.
https://github.com/penbek7-maker/muse-microstates
Hyponoia Generative Composer
Adaptive generative engine, sonic-memory architecture, computational critic, human-feedback system, and persistent learning framework.
https://github.com/penbek7-maker/hyponoia-generative-composer
Hyponoia is developed by Penelope Bekiari as part of ongoing doctoral artistic research in electroacoustic composition, biointeractive performance, affective technologies, and human–AI musical interaction.
Samothrace, with its rich cultural and natural heritage, offers a unique opportunity for exploring the relationship between people, landscape, and sound. Within the framework of the artistic residency, my goal was to highlight the sonic identity of the island through field recordings, testimonies from residents, and the composition of contemporary soundscapes.
This process is linked to the concept of "psychogeography," examining how the sounds surrounding communities shape their perception of space and identity.
Special emphasis was placed on the natural diversity of the island, the relationship between the mountain and the sea, as well as the ancient heritage of the Cabirian Mysteries. Additionally, the traditional music of Samothrace and the stories that accompany it form an integral part of the local collective memory, and for this reason, it was incorporated and embedded into the soundscape, from which it naturally originates.
Artist in Residence at the Greek National Opera as part of the National Recovery and Resilience Plan, funded by the European Commission – Next Generation EU
"Lalē (meaning “pebbles”) derives from the Greek verb lalō (“to speak”) and refers to the small stones carried by the waves along the shoreline, whose movement creates a subtle and ever-changing sonic landscape. Rooted in the coastal environment and cultural traditions of Mani, Greece, the work explores the sea as both a physical and symbolic space of memory, connection, and transformation. In its current form, Lalē is conceived as a performance for voice, piano, cello, fixed media, and live electronics. The vocalist occupies a central role, shaping and influencing the electronic processing of the voice in real time, while the instrumental performers respond and adapt to its evolving gestures and expressive qualities. Drawing inspiration from the vocal tradition of the Maniot moirologia (lament songs), the work engages with practices of oral transmission, collective memory, and emotional expression that have long formed part of Mediterranean cultural life. The performance incorporates excerpts and imagery from John Masefield’s poem Sea Fever, whose longing for the sea and the freedom of voyage resonate with the work’s broader themes of movement, belonging, and human interconnectedness. Here, the sea becomes a metaphorical space that transcends geographical boundaries, linking people, histories, and cultures across Europe through shared experiences of migration, memory, loss, and renewal. Combining spoken word, song, electroacoustic textures, environmental recordings, and live electronic processing, Lalē brings together sound, poetry, and tradition within a contemporary performance framework. The work seeks not only to evoke the sonic identity of Mani but also to reflect on the enduring capacity of the sea to unite distant places, voices, and communities.
This selected work, 'Andromeda,' is featured in the MA/IN festival and has been also selected as a poster presentation at the Advances in Data Science and AI Conference, University of Manchester and as a paper and piece in SMC Porto 2024. The live performance introduces a multimedia composition for flute and live sound processing, inspired by Euripides' tragedy.
It integrates biofeedback data from the flutist's body sensors, synchronizing sound processes with physiological responses. 'Andromeda' explores the fusion of theatrical conventions and performer emotions, leveraging biometric data to enrich narrative, development, and aesthetics. It examines real-time physiological data's role in enhancing interactive performance, bridging technology and musical expression realms. Through biometric analysis, it delves into themes of technological posthumanism, employing sensors like galvanic skin response, heart rate, and respiratory sensors. Symbolizing love, 'Andromeda' portrays the protagonist's struggle amidst echoes, waves, and wind, as she appeals to the night goddess for freedom to love Perseus.
The glass animals' sounds transform into fragmented pieces, producing noises amplified by imagination. The flowing sonic forms reflect Laura's various mental states, existing between the material world and the conceptual framework of memory.
Presented at - Stavros Niarchos Foundation Cultural Center 2024 and SAG (Leicester)
The thematic essence revolves around a synopsis of a day in Athens, spanning from the invigorating morning energies to the tranquil shades of night. The violin, with its roots in Greek traditional rhythms and folk songs, acts as a conduit, bridging ancient Greek musical texts, traditional rhythms, and contemporary technology for audio data processing. This harmonious interplay seeks to unite the echoes of the past with the rhythms of the present, offering a musical journey that transcends temporal boundaries.
Presented at Greek and international festivals including Mantis, 41st International Confederation of Electroacoustic Music Meeting, HELMCA, Salle Molière – Opéra Comédie of Montpellier - Maison des Arts Sonores – KLANG
Suffocating cities at an increasing pace. "Certain circles" explores the path from serenity to discontentment, the modern sonic landscape as a linear narration of interpersonal relationships. Echoes of a city, texts, and computer-generated sounds turn into mechanical sound gestures, struggling to fit into given spacetime. Inspired by Delia Derbyshire's "Inventions for Radio", this 8-channel piece seeks a conversation with the contemporary perspective of time and information.
It was presented at John Rylands Library and Research Institute- University of Manchester, memorial to Delia Darbishire.
Proximity | by Penelope Bekiari & Mari Lesteberg, for voices & live electronics (Super Collider / Pure Data, loopstation)
The nature and language of two different countries are expressed and blended together to tell a story about communication. The words of two poems are used to learn how to talk together and to explore the importance of silence, sound and proximity. Poems expressed in the piece: Gunvor Hofmo - Dette er Våren, "Πληγωμένη Άνοιξη [Wounded Spring]"- Miltos Sachtouris
It was presented at the Ultima Music Festival.
