update

install.sh

@@ -53,10 +53,10 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
   cd "${thirdparty_dir}"
 
   # install onnxruntime
+  # https://onnxruntime.ai/docs/install/
+  # https://github.com/microsoft/onnxruntime/releases
   wget https://github.com/microsoft/onnxruntime/releases/download/v1.20.1/onnxruntime-linux-x64-1.20.1.tgz
   tar -zxvf onnxruntime-linux-x64-1.20.1.tgz
-#  export C_INCLUDE_PATH=$C_INCLUDE_PATH:$(pwd)/onnxruntime-linux-x64-1.20.1/include
-#  export LIBRARY_PATH=$LIBRARY_PATH:$(pwd)/onnxruntime-linux-x64-1.20.1/lib
 
 fi
 

main.py

@@ -46,15 +46,16 @@ def run_nx_vad(filename: str, silence_time: float = 0.3, longest_activate: float
     cmd = "vad_bins/nx_vad --filename {} --silence_time {} --longest_activate {}".format(
         filename, silence_time, longest_activate
     )
-    vad_result = Command.popen(cmd)
+    raw_vad_result = Command.popen(cmd)
 
-    pattern = "(\\d+)[\r\n]VadFlagPrepare[\r\n](\\d+)[\r\n]VadFlagSpeaking(?:[\r\n](?:\\d+)[\r\n]VadFlagPause[\r\n](?:\\d+)[\r\n]VadFlagSpeaking)?[\r\n](\\d+)[\r\n]VadFlagNoSpeech"
+    pattern = "(\\d+)[\r\n]VadFlagPrepare[\r\n](?:\\d+)[\r\n]VadFlagSpeaking(?:[\r\n](?:\\d+)[\r\n]VadFlagPause[\r\n](?:\\d+)[\r\n]VadFlagSpeaking)?[\r\n](\\d+)[\r\n]VadFlagNoSpeech"
 
-    vad_timestamps = re.findall(pattern, vad_result, flags=re.DOTALL)
+    vad_timestamps = re.findall(pattern, raw_vad_result, flags=re.DOTALL)
+    vad_timestamps = [(float(start), float(end)) for start, end in vad_timestamps]
 
     vad_timestamps: str = json.dumps(vad_timestamps, ensure_ascii=False, indent=2)
 
-    return vad_timestamps
+    return raw_vad_result, vad_timestamps
 
 
 def run_silero_vad(filename: str, silence_time: float = 0.3, longest_activate: float = 3.0) -> str:
@@ -65,9 +66,16 @@ def run_silero_vad(filename: str, silence_time: float = 0.3, longest_activate: f
     cmd = "vad_bins/silero {}".format(
         filename
     )
-    vad_result = Command.popen(cmd)
+    raw_vad_result = Command.popen(cmd)
 
-    return vad_result
+    pattern = ".*?speech starts at (.+?)s[\r\n].*?speech ends at (.+?)s"
+
+    vad_timestamps = re.findall(pattern, raw_vad_result, flags=re.DOTALL)
+    vad_timestamps = [(float(start), float(end)) for start, end in vad_timestamps]
+
+    vad_timestamps: str = json.dumps(vad_timestamps, ensure_ascii=False, indent=2)
+
+    return raw_vad_result, vad_timestamps
 
 
 def shell(cmd: str):

test.py

@@ -0,0 +1,27 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import re
+
+string = """
+2024/12/11 06:10:03 speech starts at 4.45s
+2024/12/11 06:10:03 speech ends at 4.96s
+2024/12/11 06:10:03 speech starts at 9.63s
+2024/12/11 06:10:03 speech ends at 9.92s
+2024/12/11 06:10:03 speech starts at 20.58s
+2024/12/11 06:10:03 speech ends at 21.02s
+2024/12/11 06:10:03 speech starts at 27.36s
+2024/12/11 06:10:03 speech ends at 27.61s
+2024/12/11 06:10:03 speech starts at 27.97s
+2024/12/11 06:10:03 speech ends at 31.20s
+2024/12/11 06:10:03 speech starts at 31.55s
+2024/12/11 06:10:03 speech ends at 33.21s
+"""
+
+pattern = ".*?speech starts at (.+?)s[\r\n].*?speech ends at (.+?)s"
+
+result = re.findall(pattern, string, flags=re.DOTALL)
+print(result)
+
+
+if __name__ == "__main__":
+    pass