From f44baaa05f038affbed14d68e7805179ca7b4bf6 Mon Sep 17 00:00:00 2001
From: Daniel Wolf <dwolf@dannad.de>
Date: Thu, 29 Sep 2016 11:41:31 +0200
Subject: [PATCH] Improve noise detection heuristic

---
 src/audio/voiceActivityDetection.cpp | 8 ++++++++
 src/phoneRecognition.cpp             | 2 +-
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/audio/voiceActivityDetection.cpp b/src/audio/voiceActivityDetection.cpp
index c03983b..3da2d85 100644
--- a/src/audio/voiceActivityDetection.cpp
+++ b/src/audio/voiceActivityDetection.cpp
@@ -100,6 +100,14 @@ BoundedTimeline<void> detectVoiceActivity(const AudioClip& inputAudioClip, int m
 		}
 	}
 
+	// Shorten activities. WebRTC adds a bit of buffer at the end.
+	const centiseconds tail(5);
+	for (const auto& utterance : Timeline<void>(activity)) {
+		if (utterance.getTimeRange().getLength() > tail && utterance.getEnd() < audioLength) {
+			activity.clear(utterance.getEnd() - tail, utterance.getEnd());
+		}
+	}
+
 	logging::debugFormat("Found {} sections of voice activity: {}", activity.size(),
 		join(activity | transformed([](const Timed<void>& t) { return format("{0}-{1}", t.getStart(), t.getEnd()); }), ", "));
 
diff --git a/src/phoneRecognition.cpp b/src/phoneRecognition.cpp
index 742de17..a0717dd 100644
--- a/src/phoneRecognition.cpp
+++ b/src/phoneRecognition.cpp
@@ -288,7 +288,7 @@ Timeline<void> getNoiseSounds(TimeRange utteranceTimeRange, const Timeline<Phone
 	}
 
 	// Remove undesired elements
-	const centiseconds minSoundLength = 5_cs;
+	const centiseconds minSoundLength = 12_cs;
 	for (const auto& unknownSound : Timeline<void>(noiseSounds)) {
 		bool startsAtZero = unknownSound.getStart() == 0_cs;
 		bool tooShort = unknownSound.getTimeRange().getLength() < minSoundLength;