From 37d6930561f4e798e76f81c37d912906c97b6bea Mon Sep 17 00:00:00 2001
From: pointerhacker <145901472+pointerhacker@users.noreply.github.com>
Date: Wed, 20 Nov 2024 18:39:05 +0800
Subject: [PATCH] =?UTF-8?q?fix:=E6=B7=BB=E5=8A=A0=E9=9F=B3=E9=A2=91?=
 =?UTF-8?q?=E7=A0=81=E7=8E=87=E9=AA=8C=E8=AF=81=E4=BB=A5=E9=98=B2=E6=AD=A2?=
 =?UTF-8?q?=E6=A8=A1=E5=9E=8B=E5=BC=82=E5=B8=B8=20(#2219)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Co-authored-by: zhaochaojin <zhaochaojin@didiglobal.com>
---
 funasr/utils/load_utils.py | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.py
index 37b3dbfbf..aabf5d9f3 100644
--- a/funasr/utils/load_utils.py
+++ b/funasr/utils/load_utils.py
@@ -1,6 +1,7 @@
 import os
 import torch
 import json
+from io import BytesIO
 import torch.distributed as dist
 import numpy as np
 import kaldiio
@@ -9,6 +10,7 @@ import torchaudio
 import time
 import logging
 from torch.nn.utils.rnn import pad_sequence
+from pydub import AudioSegment
 
 try:
     from funasr.download.file import download_from_url
@@ -136,6 +138,7 @@ def load_audio_text_image_video(
 
 
 def load_bytes(input):
+    input = validate_frame_rate(input)
     middle_data = np.frombuffer(input, dtype=np.int16)
     middle_data = np.asarray(middle_data)
     if middle_data.dtype.kind not in "iu":
@@ -150,6 +153,31 @@ def load_bytes(input):
     array = np.frombuffer((middle_data.astype(dtype) - offset) / abs_max, dtype=np.float32)
     return array
 
+def validate_frame_rate(
+    input,
+    fs: int = 16000,
+):
+    
+    # 将文件读取为字节流
+    byte_data = BytesIO(input)
+
+    # 使用 pydub 加载音频
+    audio = AudioSegment.from_file(byte_data)
+
+    # 确保采样率为 16000 Hz
+    if audio.frame_rate != fs:
+        audio = audio.set_frame_rate(fs)
+
+        # 将重新采样后的音频导出为字节流
+        output = BytesIO()
+        audio.export(output, format="wav")
+        output.seek(0)
+
+        # 获取重新采样后的字节流数据
+        input = output.read()
+        
+    return input
+
 
 def extract_fbank(data, data_len=None, data_type: str = "sound", frontend=None, **kwargs):
     if isinstance(data, np.ndarray):