From 2774fb8379238600a9efc410f049575f4c003b6b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Felix=20F=C3=B6rtsch?= <mail@felixfoertsch.de>
Date: Thu, 19 Feb 2026 12:36:29 +0100
Subject: [PATCH] add per-channel min-max normalization for ASR features

---
 VorleserMac/Services/KokoroPipeline.swift | 33 +++++++++++++++++++++++
 1 file changed, 33 insertions(+)

diff --git a/VorleserMac/Services/KokoroPipeline.swift b/VorleserMac/Services/KokoroPipeline.swift
index 5a1de3f..f4526d3 100644
--- a/VorleserMac/Services/KokoroPipeline.swift
+++ b/VorleserMac/Services/KokoroPipeline.swift
@@ -279,6 +279,7 @@ final class KokoroPipeline {
             return channel * (shape.last ?? 0) + token
         }
 
+        // First pass: compute weighted sum
         for h in 0..<availableChannels {
             for f in 0..<frameCount {
                 var sum: Float = 0
@@ -295,6 +296,38 @@ final class KokoroPipeline {
                 output[outIndex] = NSNumber(value: sum)
             }
         }
+
+        // Second pass: per-channel min-max normalization
+        var channelMin = [Float](repeating: Float.greatestFiniteMagnitude, count: availableChannels)
+        var channelMax = [Float](repeating: -Float.greatestFiniteMagnitude, count: availableChannels)
+
+        for h in 0..<availableChannels {
+            for f in 0..<frameCount {
+                let idx = h * frameCount + f
+                let value = output[idx].floatValue
+                if value.isFinite {
+                    if value < channelMin[h] { channelMin[h] = value }
+                    if value > channelMax[h] { channelMax[h] = value }
+                }
+            }
+        }
+
+        for h in 0..<availableChannels {
+            let range = channelMax[h] - channelMin[h]
+            if range > 1e-6 {
+                for f in 0..<frameCount {
+                    let idx = h * frameCount + f
+                    let value = output[idx].floatValue
+                    if value.isFinite {
+                        let normalized = (value - channelMin[h]) / range * 2.0 - 1.0
+                        output[idx] = NSNumber(value: normalized)
+                    } else {
+                        output[idx] = NSNumber(value: Float(0))
+                    }
+                }
+            }
+        }
+
         return output
     }