feat(analyzer): add speech candidate detection for adaptive tuning

[flexiondotorg]

Summary

Add speech candidate detection to identify representative speech regions for future adaptive filter tuning. This complements the existing silence detection by providing measurements of typical speech characteristics.

Changes

Speech Detection (commit 1)

• Add SpeechRegion and SpeechCandidateMetrics data structures
• Implement interval-based speech detection after elected silence
• Score speech regions by amplitude, centroid, and entropy
• Select longest qualifying candidate for speech profiling
• Integrate detection into MeasureInput analysis pipeline
• Add diagnostic output to processing reports
• Add 21 unit tests for speech detection logic

Expanded Metrics (commit 2)

• Add all spectral metrics (mean, variance, spread, skewness, crest, flux, slope, decrease, rolloff) to silence and speech candidates
• Add loudness metrics (momentary/short-term LUFS, true/sample peak)
• Update measurement functions to populate complete metric set
• Enhance diagnostic report with organised metric groups

Design

• Speech detection runs after silence detection, searching only after the elected silence region ends
• Uses 30-second minimum duration with 2-second interruption tolerance for natural pauses
• Scoring prioritises amplitude (50%), voice-range centroid (30%), and low entropy (20%)
• Selection prefers longest duration above quality threshold (unlike silence which prefers earliest)

Testing

All existing tests pass. New tests cover:

• TestSpeechScore — 6 test cases for speech scoring
• TestFindSpeechCandidatesFromIntervals — 6 test cases for detection logic
• TestMeasureSpeechCandidateFromIntervals — 2 test cases for metrics extraction
• TestFindBestSpeechRegion — 3 test cases for candidate selection
• TestScoreSpeechCandidate — 4 test cases for candidate scoring

Implements docs/PLAN-SpeechDetection.md

[flexiondotorg]

@cubic-dev-ai Review this pull request

[cubic-dev-ai]

@cubic-dev-ai Review this pull request

@flexiondotorg I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

1 issue found across 3 files

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="internal/processor/analyzer.go">

<violation number="1" location="internal/processor/analyzer.go:2777">
P1: If no interval exists at or after `searchStart`, `startIdx` remains 0 and speech detection incorrectly searches from the beginning of the file instead of returning nil. This could detect speech within the silence region for short recordings or when silence is near the end.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}