dsh-fish-pet 0.2.4 → 0.2.6

package/assets/asr-listener.ps1

@@ -2,16 +2,16 @@
 # 一次连续采集，检测到"说完"立刻上传，不再固定切片
 param(
     [int]$MaxMs = 10000,
-    [int]$SilenceMs = 600,
+    [int]$SilenceMs = 900,
     [int]$Boost = 1,
-    [int]$SilenceGate = 700
+    [int]$SilenceGate = 600
 )
 
-# 单实例：防止耳朵双开（双开会重复识别、重复回话）
-try {
-    $earMutex = New-Object System.Threading.Mutex($false, 'Global\dsh-fish-ear-single')
-    if (-not $earMutex.WaitOne(0)) { exit 0 }
-} catch { }
+# 单实例：防止耳朵双开（双开会重复识别、重复回话）
+try {
+    $earMutex = New-Object System.Threading.Mutex($false, 'Global\dsh-fish-ear-single')
+    if (-not $earMutex.WaitOne(0)) { exit 0 }
+} catch { }
 $ErrorActionPreference = 'Stop'
 try { Add-Type -AssemblyName System.Net.Http -ErrorAction Stop } catch { }
 $DataDir = Join-Path $env:USERPROFILE '.dsh\fish-pet'
@@ -28,6 +28,7 @@ if (Test-Path $CfgPath) {
         $cj = Get-Content -LiteralPath $CfgPath -Raw | ConvertFrom-Json
         if ($cj.voice) {
             if ($cj.voice.silenceGate) { $SilenceGate = [int]$cj.voice.silenceGate }
+    if ($cj.voice.silenceMs) { $SilenceMs = [int]$cj.voice.silenceMs }
             if ($cj.voice.boost) { $Boost = [int]$cj.voice.boost }
             if ($cj.voice.engine) { $script:engine = [string]$cj.voice.engine }
             if ($cj.voice.localConf) { $script:localConf = [double]$cj.voice.localConf }
@@ -133,19 +134,25 @@ public class EarVad
                 all.AddRange(buf);
                 if (peakInt > RawPeak) RawPeak = peakInt;
                 if (peakInt < gate) silent += packetMs; else silent = 0;
-                if (silent >= silenceMs) break;
+                if (silent >= silenceMs && (total - silent) >= 1200) break;
             }
         }
         ac.Stop();
         if (!started || all.Count < rate / 2) return false;
 
-        // 轻声增益：整段峰值偏低时放大，响的句子不动
-        float maxA = 0; double sumSq = 0; int cntAll = 0;
-        for (int i = 0; i < all.Count; i++) { float a = Math.Abs(all[i]); if (a > maxA) maxA = a; sumSq += all[i] * all[i]; cntAll++; }
-        double peakInt2 = maxA * 32767.0;
-        double gain2 = 1.0;
-        if (peakInt2 > 150 && peakInt2 < 8000) { gain2 = Math.Min(12.0, 20000.0 / peakInt2); }
-
+        // 轻声增益：整段峰值偏低时放大，响的句子不动
+        float maxA = 0; double sumSq = 0; int cntAll = 0;
+        for (int i = 0; i < all.Count; i++) { float a = Math.Abs(all[i]); if (a > maxA) maxA = a; sumSq += all[i] * all[i]; cntAll++; }
+        double peakInt2 = maxA * 32767.0;
+        double gain2 = 1.0;
+        // 峰值归一化：把这段音频拉到约 70% 满幅，轻声也能被云端听清
+        if (peakInt2 > 200) {
+            gain2 = 22936.0 / peakInt2;
+            if (gain2 < 1.0) gain2 = 1.0;
+            if (gain2 > 35.0) gain2 = 35.0;
+            // 极轻的声音额外再提一点（>35 倍会明显失真，所以封顶）
+        }
+
         double step = (double)rate / 16000.0;
         int framesIn = all.Count / ch;
         int outCount = (int)(framesIn / step);
@@ -263,6 +270,12 @@ $script:lastFishCheck = Get-Date
         if (-not $key) { Start-Sleep -Seconds 2; continue }
 
         $ok = [EarVad]::Capture($MaxMs, $SilenceGate, $SilenceMs, $Boost, 2)
+
+        # 录音过程中鱼开口说话（或刚说完）→ 这段是自己的声音，丢掉
+        $talkF = Join-Path $DataDir 'asr.talking'
+        if (Test-Path $talkF) { continue }
+        $pauseF = Join-Path $DataDir 'asr.pause'
+        if ((Test-Path $pauseF) -and (((Get-Date) - (Get-Item $pauseF).LastWriteTime).TotalSeconds -lt 2.0)) { continue }
         if (-not $ok) {
             Start-Sleep -Milliseconds 80
             continue
@@ -290,23 +303,23 @@ $script:lastFishCheck = Get-Date
             }
         }
         if (-not $text -and $engine -ne 'local') {
-            if ($null -eq $script:http) {
-                $script:http = New-Object System.Net.Http.HttpClient
-                $script:http.Timeout = [TimeSpan]::FromSeconds(30)
-                $script:http.DefaultRequestHeaders.Authorization = New-Object System.Net.Http.Headers.AuthenticationHeaderValue('Bearer', $key)
-            }
-            $json = ''
-            try {
-                $mp = New-Object System.Net.Http.MultipartFormDataContent
-                $bytes = [IO.File]::ReadAllBytes($wav)
-                $bc = New-Object System.Net.Http.ByteArrayContent((,$bytes))
-                $bc.Headers.ContentType = [System.Net.Http.Headers.MediaTypeHeaderValue]::Parse('audio/wav')
-                $mp.Add($bc, 'file', 'a.wav')
-                $mp.Add((New-Object System.Net.Http.StringContent('Qwen/Qwen3-ASR-1.7B')), 'model')
-                $mp.Add((New-Object System.Net.Http.StringContent('zh')), 'language')
-                $resp = $script:http.PostAsync('https://api.siliconflow.cn/v1/audio/transcriptions', $mp).GetAwaiter().GetResult()
-                $json = $resp.Content.ReadAsStringAsync().GetAwaiter().GetResult()
-                try { $resp.Dispose(); $mp.Dispose() } catch { }
+            if ($null -eq $script:http) {
+                $script:http = New-Object System.Net.Http.HttpClient
+                $script:http.Timeout = [TimeSpan]::FromSeconds(30)
+                $script:http.DefaultRequestHeaders.Authorization = New-Object System.Net.Http.Headers.AuthenticationHeaderValue('Bearer', $key)
+            }
+            $json = ''
+            try {
+                $mp = New-Object System.Net.Http.MultipartFormDataContent
+                $bytes = [IO.File]::ReadAllBytes($wav)
+                $bc = New-Object System.Net.Http.ByteArrayContent((,$bytes))
+                $bc.Headers.ContentType = [System.Net.Http.Headers.MediaTypeHeaderValue]::Parse('audio/wav')
+                $mp.Add($bc, 'file', 'a.wav')
+                $mp.Add((New-Object System.Net.Http.StringContent('Qwen/Qwen3-ASR-1.7B')), 'model')
+            # 不指定 language，交给模型自动判断中英文（指定 zh 时英文会被转成奇怪文字）
+                $resp = $script:http.PostAsync('https://api.siliconflow.cn/v1/audio/transcriptions', $mp).GetAwaiter().GetResult()
+                $json = $resp.Content.ReadAsStringAsync().GetAwaiter().GetResult()
+                try { $resp.Dispose(); $mp.Dispose() } catch { }
             } catch { try { Add-Content -LiteralPath $ErrFile -Value ((Get-Date -Format 'HH:mm:ss') + '  HTTP ' + $_.Exception.Message) -Encoding UTF8 } catch { } }
             if ($json) {
                 try { $obj = $json | ConvertFrom-Json; $text = [string]$obj.text; $via = 'cloud' } catch { }
@@ -315,9 +328,14 @@ $script:lastFishCheck = Get-Date
         Remove-Item -LiteralPath $wav -Force -ErrorAction SilentlyContinue
         Add-Content -LiteralPath $LevelFile -Value ((Get-Date -Format 'HH:mm:ss') + "`tengine=" + $via + "`ttext=" + $text) -Encoding UTF8
         if ($text) { $text = $text.Trim() }
-        if (-not $text -and [EarVad]::RawPeak -ge 800) {
-            Add-Content -LiteralPath $OutFile -Value ((Get-Date -Format 'HH:mm:ss') + "`t?`t(没听清)") -Encoding UTF8
-        }
+        # 过滤乱语种：泰文/日文假名/西里尔/阿拉伯等说明识别跑偏了，当没听清
+        if ($text -and ($text -match '[\u0e00-\u0eff\u3040-\u30ff\u0400-\u04ff\u0600-\u06ff]')) {
+            Add-Content -LiteralPath $OutFile -Value ((Get-Date -Format 'HH:mm:ss') + "`t?`t(识别跑偏：" + $text + ")") -Encoding UTF8
+            $text = ''
+        }
+        if (-not $text -and [EarVad]::RawPeak -ge 800) {
+            Add-Content -LiteralPath $OutFile -Value ((Get-Date -Format 'HH:mm:ss') + "`t?`t(没听清)") -Encoding UTF8
+        }
         if ($text -and ($text -match '[\u4e00-\u9fa5]' -or $text.Length -ge 4)) {
             Add-Content -LiteralPath $OutFile -Value ((Get-Date -Format 'HH:mm:ss') + "`t1`t" + $text) -Encoding UTF8
         }