From 87ab7d4b9cdf85b8fab4224195b21997a17307df Mon Sep 17 00:00:00 2001
From: admin <572701190@qq.com>
Date: Sat, 2 May 2026 06:20:48 +0800
Subject: [PATCH] Stabilize speech audio capture

- Keep Web Audio source and processor nodes in refs so microphone processing is not lost during speech recognition.

- Explicitly resume the AudioContext before sending PCM frames to the speech WebSocket proxy.

- Disconnect audio nodes and clear speech state when listening stops or the WebSocket closes.

- Show a clear prompt when audio frames were captured but no recognizable text was returned.

- Update progress documentation for the speech capture hardening.
---
 docs/progress.md           |  1 +
 src/pages/ReportEditor.tsx | 39 +++++++++++++++++++++++++++++++++++++-
 2 files changed, 39 insertions(+), 1 deletion(-)

diff --git a/docs/progress.md b/docs/progress.md
index 49240e6..ee567ac 100644
--- a/docs/progress.md
+++ b/docs/progress.md
@@ -93,3 +93,4 @@
 | 2026-05-02 | 模板 HTML 导出包补充模板字段和字段管理设置，导入时恢复字段库元数据。 |
 | 2026-05-02 | 修复报告编辑器加载已有 AI 区域后下拉栏初始显示“无可用 AI 区域”的问题。 |
 | 2026-05-02 | 调整抽帧百分比为两位小数保序保存；自动截图按时间顺序执行，自动插入按配置顺序执行。 |
+| 2026-05-02 | 加固报告编辑器语音采集，保留 Web Audio 节点引用、显式恢复 AudioContext，并在无识别文本时给出提示。 |
diff --git a/src/pages/ReportEditor.tsx b/src/pages/ReportEditor.tsx
index a8a7c54..97e752b 100644
--- a/src/pages/ReportEditor.tsx
+++ b/src/pages/ReportEditor.tsx
@@ -105,6 +105,11 @@ export default function ReportEditor() {
   const xfWsRef = useRef<WebSocket | null>(null);
   const xfAudioContextRef = useRef<AudioContext | null>(null);
   const xfMediaStreamRef = useRef<MediaStream | null>(null);
+  const xfAudioSourceRef = useRef<MediaStreamAudioSourceNode | null>(null);
+  const xfAudioProcessorRef = useRef<ScriptProcessorNode | null>(null);
+  const xfSpeechFrameCountRef = useRef(0);
+  const xfSpeechTextReceivedRef = useRef(false);
+  const xfSpeechUserStoppedRef = useRef(false);
   const [quickPrompts, setQuickPrompts] = useState<string[]>([
     '请完善报告内容', '请对内容做如下修改：'
   ]);
@@ -1143,6 +1148,17 @@ export default function ReportEditor() {
   const toggleListening = async () => {
     // 专门提取一个彻底关闭物理麦克风的函数
     const stopMicrophone = () => {
+      if (xfAudioProcessorRef.current) {
+        try {
+          xfAudioProcessorRef.current.onaudioprocess = null;
+          xfAudioProcessorRef.current.disconnect();
+        } catch {}
+        xfAudioProcessorRef.current = null;
+      }
+      if (xfAudioSourceRef.current) {
+        try { xfAudioSourceRef.current.disconnect(); } catch {}
+        xfAudioSourceRef.current = null;
+      }
       if (xfAudioContextRef.current) {
         try { xfAudioContextRef.current.close(); } catch {}
         xfAudioContextRef.current = null;
@@ -1155,6 +1171,7 @@ export default function ReportEditor() {
 
     if (isListening) {
       setIsListening(false);
+      xfSpeechUserStoppedRef.current = true;
       stopMicrophone();
 
       if (xfWsRef.current && xfWsRef.current.readyState === WebSocket.OPEN) {
@@ -1182,6 +1199,9 @@ export default function ReportEditor() {
 
       const ws = new WebSocket(getSpeechIatWebSocketUrl());
       xfWsRef.current = ws;
+      xfSpeechFrameCountRef.current = 0;
+      xfSpeechTextReceivedRef.current = false;
+      xfSpeechUserStoppedRef.current = false;
       let frameStatus = 0;
 
       ws.onopen = async () => {
@@ -1190,8 +1210,13 @@ export default function ReportEditor() {
           xfMediaStreamRef.current = stream;
           const audioContext = new AudioContextClass({ sampleRate: 16000 });
           xfAudioContextRef.current = audioContext;
+          if (audioContext.state === 'suspended') {
+            await audioContext.resume();
+          }
           const source = audioContext.createMediaStreamSource(stream);
           const processor = audioContext.createScriptProcessor(4096, 1, 1);
+          xfAudioSourceRef.current = source;
+          xfAudioProcessorRef.current = processor;
 
           processor.onaudioprocess = (e) => {
             if (ws.readyState !== WebSocket.OPEN || !xfAudioContextRef.current) return;
@@ -1200,6 +1225,7 @@ export default function ReportEditor() {
             const base64Audio = arrayBufferToBase64(pcmBuffer);
             const frame: any = { data: { status: frameStatus, format: 'audio/L16;rate=16000', encoding: 'raw', audio: base64Audio } };
             ws.send(JSON.stringify(frame));
+            xfSpeechFrameCountRef.current += 1;
             frameStatus = 1;
           };
 
@@ -1227,6 +1253,7 @@ export default function ReportEditor() {
             let seg = '';
             for (const w of jsonData.data.result.ws) { if (w.cw?.[0]?.w) seg += w.cw[0].w; }
             if (seg) {
+              xfSpeechTextReceivedRef.current = true;
               setChatInput(prev => prev + seg);
             }
           }
@@ -1240,7 +1267,17 @@ export default function ReportEditor() {
       };
 
       ws.onerror = () => { alert('讯飞语音连接失败，请确认已登录且超级管理员已配置语音参数'); setIsListening(false); stopMicrophone(); };
-      ws.onclose = () => { setIsListening(false); stopMicrophone(); };
+      ws.onclose = () => {
+        const shouldExplainNoText = xfSpeechUserStoppedRef.current
+          && xfSpeechFrameCountRef.current > 0
+          && !xfSpeechTextReceivedRef.current;
+        setIsListening(false);
+        stopMicrophone();
+        xfWsRef.current = null;
+        if (shouldExplainNoText) {
+          alert('语音听写已结束，但讯飞没有返回可用文字。请确认麦克风输入音量正常，并尽量使用普通话靠近麦克风重试。');
+        }
+      };
     } catch (e: any) {
       alert('讯飞语音初始化失败: ' + e.message);
     }