Room Acoustics Measurement Study

Skills Demonstrated

• Signal processing: Farina analytical inverse-filter deconvolution, Lundeby noise-floor truncation, Schroeder backward integration with Chu noise correction.
• MATLAB pipeline development with synthetic ground-truth validation using the normalized misalignment metric.
• Applied statistics: blocked two-way ANOVA, Tukey–Kramer post-hoc testing, composite scoring with linear error propagation for confidence intervals.
• Experimental design and field measurement in accordance with ISO 3382-1/2 and ANSI S12.60.
• Technical writing: formal engineering report with figure design, tabulated results, and a rigorous error/limitations discussion.

Problem & Context

• Course: ME 4505 term project in the Department of Mechanical and Industrial Engineering at Northeastern University (Spring 2026).
• Problem: no public acoustic characterization existed for Northeastern's teaching spaces, yet students anecdotally perceive a wide range of acoustic quality across rooms.
• Goal: produce an objective ranking of a representative set of rooms using ISO 3382 parameters, identify which rooms meet the ANSI S12.60 35 dB(A) background-noise limit for core learning spaces, and deliver a reusable measurement pipeline for future campus work.

Process & Constraints

• Measured nine rooms spanning small classrooms, medium lecture halls, and large auditoria (EXP 204, ISEC 102, Snell 168, Robinson 109, EV 002, Mugar 201, WVG 108, WVF 020, Shillman 215).
• Excitation: 10-second exponential sine sweep (20 Hz – 20 kHz) generated in MATLAB; captured at 48 kHz / 24-bit in Audacity.
• Per-room protocol: six recordings at two source positions crossed with three receiver positions (front, middle, back) — 54 measurements total.
• Equipment: Jounivo JV-601 USB condenser microphone, JBL Flip 5 portable loudspeaker, ASUS Zephyrus G14 laptop; iPhone-based sound level meter for A-weighted background noise.
• Analysis pipeline (MATLAB): Farina analytical inverse filter for deconvolution, Lundeby iterative noise-floor truncation, Schroeder backward integration with Chu noise correction, then ISO 3382 parameter extraction.
• Validation: pipeline verified against a synthetic ground truth (known impulse response convolved with the ESS plus additive white noise), achieving a normalized misalignment of −11.4 dB with no regularization parameter required.
• Cross-room statistics: blocked two-way additive ANOVA with source-receiver position as a blocking factor, followed by Tukey–Kramer post-hoc comparisons across all 36 room pairs.
• Composite speech-quality score: weighted combination of D50 (40%), a T20 penalty against a 0.7 s target (25%), and LAeq (35%), with 95% confidence intervals from linear error propagation.
• Constraints: cardioid microphone (in place of the ISO 3382-1 omnidirectional reference) and a single powered loudspeaker (in place of a dodecahedral source) introduced consistent directivity biases — held constant across rooms and documented in the report so relative rankings remain valid.

Results & Impact

• Produced the first public acoustic characterization of nine Northeastern teaching spaces.
• Observed parameter ranges: T20 from 0.40 s (Snell 168) to 1.00 s (EXP 204); D50 from 0.78 to 0.93; background noise LAeq from 30.5 to 45.3 dB(A).
• Two-way ANOVA found room effects highly significant for all seven ISO 3382 parameters (p < 10⁻⁶).
• Composite ranking: ISEC 102 scored highest (+0.54) and EXP 204 lowest (−0.89), with non-overlapping 95% confidence intervals at the extremes of the distribution.
• Identified that most measured rooms exceed the ANSI S12.60 35 dB(A) background-noise limit for core learning spaces even when unoccupied.
• Flagged EXP 204 as a strong candidate for acoustic treatment on the basis of its ~1.0 s reverberation time (roughly double the 0.4–0.8 s range recommended for speech).
• Delivered a reusable MATLAB measurement and analysis pipeline — including synthetic ground-truth validation — released as v1.0 on GitHub for future campus measurement work.

Team & Contribution

• Three-person team project.
• Personally owned MATLAB measurement and analysis pipeline development, ISO 3382 parameter extraction, cross-room statistical analysis, and written report authorship.
• Field measurements conducted as a three-person crew with fixed per-member roles (microphone/laptop, loudspeaker/source placement, LAeq reading) held constant across all nine rooms to reduce operator variability.

Media

Full project report embedded below. The complete measurement and analysis pipeline (MATLAB source, raw recordings, processed impulse responses, and cross-room comparison outputs) is available on GitHub at github.com/aldernadon/room-acoustics.