Adaptive, Low-Latency Collaborative Music Streaming System for Small Ensembles

Thesis Abstract

The proliferation of remote musical collaboration has highlighted a persistent latency bottleneck that impedes real-time ensemble performance, particularly for small groups relying on live interaction. This study addresses the problem of maintaining tight synchrony and perceptual coherence in adaptive, low-latency collaborative music streaming across heterogeneous network conditions and device capabilities. The aim is to design, implement, and evaluate a streaming system that dynamically optimizes latency, jitter, and packet loss concealment while preserving musical expressivity and user experience for small ensembles of 3–6 performers. Specific objectives are (1) to develop an adaptive streaming architecture that jointly optimizes transport, encoding, and synchronization control; (2) to implement a modular prototype supporting standard digital audio workstations and mobile devices with end-to-end latency under 20 ms for intra-team collaboration under typical home-network conditions; (3) to evaluate perceptual synchrony, latency tolerance, and musical quality using both objective metrics and listener-based assessments; (4) to analyze the impact of network variability on ensemble cohesion and individual performers’ performance quality; and (5) to formulate design guidelines for scalable, ceremony-free low-latency collaboration in educational and small-ensemble contexts. The methodology adopts an action-research oriented, mixed-methods design aligned with the constructivist-interpretivist paradigm to iteratively develop and evaluate the system. The population comprises professional and semi-professional instrumentalists engaged in small-ensemble practice, with a target sample of 60 participants across three metropolitan conservatories. A stratified purposive sample (n=20) will be drawn for quantitative testing of latency and synchronization under controlled network emulations, and a qualitative subsample (n=40) will participate in collaborative sessions to gather in-depth perceptual and workflow data. Data collection instruments include a custom low-latency streaming testbed with adjustable network conditions, spectrographic and time-domain latency analyzers, standardized perceptual synchrony scales, and a semi-structured observer protocol. Reliability and validity will be enhanced through repeated-measures testing, calibration sessions, and triangulation across instrumented metrics and expert evaluations. Data analysis will employ a combination of regression analysis to model the relationship between network jitter and perceived synchrony, ANOVA to compare performance across latency configurations and encoding schemes, and mixed-effects modeling to account for participant-level variance. The qualitative component will utilize thematic analysis of interview transcripts and session recordings, grounded in the sociomusical framework and the theory of perceptual unity, to interpret practitioners’ experiences of latency, control, and collaborative flow. A conceptual model will integrate the adaptive control loop, end-to-end latency, and perceptual coherence as antecedents of ensemble cohesion. The expected findings indicate that dynamic adaptation of encoding rate, packet pacing, and clock synchronization can reduce perceived latency while maintaining musical quality, achieving a statistically significant improvement in synchrony scores (p<0.05) compared with fixed-configuration baselines. It is anticipated that a latency range of 12–20 ms intra-ensemble, coupled with jitter compensation and predictive buffering, will support near-seamless coordination for most musical styles represented in the sample. Differences across instrument families (e.g., rhythm section vs. melodic instruments) are expected to emerge, with rhythm-driven parts showing greater tolerance for latency variance than melody-centric lines. The study contributes to knowledge by (i) proposing a holistic, architecture-level design for low-latency collaborative music streaming; (ii) providing empirical evidence on the trade-offs between latency, audio quality, and synchronization; and (iii) offering practical guidelines for educators and ensemble practitioners implementing remote collaboration platforms. The theoretical contribution integrates Teleology of Musical Action and Synchrony Theory to frame how adaptive control strategies influence performers’ perceived unity and embodiment within networked ensembles. The main conclusion will articulate that an adaptive, end-to-end streaming system with real-time feedback on network conditions and perceptual synchrony can substantially preserve ensemble cohesion in small groups without requiring specialized hardware. Recommendations will include refinements to encoding heuristics, cross-layer synchronization protocols, and user-interface features that facilitate performer awareness. Directions for future research will address scalability to larger ensembles, integration with cloud-based rehearsal spaces, and long-term studies of learning outcomes in remote ensemble training.