Multi-speaker generalization study (beyond n=40 set)

[awais-de]

Parent: none -- standalone extension

Key finding

Pending evaluation

Hypothesis

n=40 CI numbers (5.89 kbps, PESQ 1.258, STOI 0.792) hold as n increases across a more diverse speaker set.

Execution checklist

• Select n>=100 speaker subset
• Run same protocol as Evaluate Phase G on n=40 speakers — kbps and STOI with 95% CI #10/Script: eval_confidence_intervals.py — n=40 CI evaluation #13

Results

Metric	n>=100	n=40 baseline
kbps	TBD	5.89
PESQ-WB	TBD	1.258
STOI	TBD	0.792

Open

• Check for subgroup effects (accent/recording condition), not just mean shift

Baseline corrected 2026-08-10 (n=40 genuine — see #10).

[awais-de]

Deferred — out of scope for the current thesis timeline.