From 1d1ef01b4e23630a99a3be7e9d1dce9550a793e9 Mon Sep 17 00:00:00 2001
From: yhliang <68215459+yhliang-aslp@users.noreply.github.com>
Date: 星期四, 11 五月 2023 16:26:24 +0800
Subject: [PATCH] Merge branch 'main' into dev_smohan

---
 funasr/runtime/onnxruntime/src/audio.cpp |   42 +++++++++++++++++++++++++++++++++++++++---
 1 files changed, 39 insertions(+), 3 deletions(-)

diff --git a/funasr/runtime/onnxruntime/src/audio.cpp b/funasr/runtime/onnxruntime/src/audio.cpp
index 8f46a4f..6d63d67 100644
--- a/funasr/runtime/onnxruntime/src/audio.cpp
+++ b/funasr/runtime/onnxruntime/src/audio.cpp
@@ -11,6 +11,7 @@
 
 using namespace std;
 
+namespace funasr {
 // see http://soundfile.sapp.org/doc/WaveFormat/
 // Note: We assume little endian here
 struct WaveHeader {
@@ -235,6 +236,24 @@
     is.read(reinterpret_cast<char *>(&header), sizeof(header));
     if(!is){
         LOG(ERROR) << "Failed to read " << filename;
+        return false;
+    }
+
+    if (!header.Validate()) {
+        return false;
+    }
+
+    header.SeekToDataChunk(is);
+    if (!is) {
+        return false;
+    }
+    
+    if (!header.Validate()) {
+        return false;
+    }
+
+    header.SeekToDataChunk(is);
+    if (!is) {
         return false;
     }
     
@@ -496,7 +515,7 @@
     delete frame;
 }
 
-void Audio::Split(Model* recog_obj)
+void Audio::Split(OfflineStream* offline_stream)
 {
     AudioFrame *frame;
 
@@ -507,7 +526,7 @@
     frame = NULL;
 
     std::vector<float> pcm_data(speech_data, speech_data+sp_len);
-    vector<std::vector<int>> vad_segments = recog_obj->VadSeg(pcm_data);
+    vector<std::vector<int>> vad_segments = (offline_stream->vad_handle)->Infer(pcm_data);
     int seg_sample = MODEL_SAMPLE_RATE/1000;
     for(vector<int> segment:vad_segments)
     {
@@ -519,4 +538,21 @@
         frame_queue.push(frame);
         frame = NULL;
     }
-}
\ No newline at end of file
+}
+
+
+void Audio::Split(VadModel* vad_obj, vector<std::vector<int>>& vad_segments)
+{
+    AudioFrame *frame;
+
+    frame = frame_queue.front();
+    frame_queue.pop();
+    int sp_len = frame->GetLen();
+    delete frame;
+    frame = NULL;
+
+    std::vector<float> pcm_data(speech_data, speech_data+sp_len);
+    vad_segments = vad_obj->Infer(pcm_data);
+}
+
+} // namespace funasr
\ No newline at end of file

--
Gitblit v1.9.1