From 5c04f7bfb84a1f2f3b307d824a1355c9c8d30a20 Mon Sep 17 00:00:00 2001
From: goddamnVincent <84380030+goddamnVincent@users.noreply.github.com>
Date: Sun, 8 Dec 2024 11:17:15 +0800
Subject: [PATCH] 'try to fix 'compute_fbank_kespeech_splits.py: error:
 unrecognized arguments: --speed-perturb true'' (#1812)

---
 .../ASR/local/compute_fbank_kespeech_dev_test.py    | 12 +++++++++++-
 .../ASR/local/compute_fbank_kespeech_splits.py      | 13 +++++++++++++
 2 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_dev_test.py b/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_dev_test.py
index 6f75dbfa49..5e169e8949 100755
--- a/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_dev_test.py
+++ b/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_dev_test.py
@@ -52,13 +52,19 @@ def get_parser():
         default=80,
         help="""The number of mel bins for Fbank""",
     )
-
     parser.add_argument(
         "--whisper-fbank",
         type=str2bool,
         default=False,
         help="Use WhisperFbank instead of Fbank. Default: False.",
     )
+    parser.add_argument(
+        "--speed-perturb",
+        type=str2bool,
+        default=False,
+        help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
+    )
+
     return parser
 
 
@@ -104,6 +110,10 @@ def compute_fbank_kespeech_dev_test(args):
             keep_overlapping=False, min_duration=None
         )
 
+        if args.speed_perturb:
+            cut_set = (
+                    cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
+            )
         logging.info("Computing features")
         cut_set = cut_set.compute_and_store_features_batch(
             extractor=extractor,
diff --git a/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_splits.py b/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_splits.py
index c398411f6e..6bb8af0d6a 100755
--- a/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_splits.py
+++ b/egs/multi_zh-hans/ASR/local/compute_fbank_kespeech_splits.py
@@ -106,6 +106,14 @@ def get_parser():
         default=False,
         help="Use WhisperFbank instead of Fbank. Default: False.",
     )
+
+    parser.add_argument(
+        "--speed-perturb",
+        type=str2bool,
+        default=False,
+        help="Enable 0.9 and 1.1 speed perturbation for data augmentation. Default: False.",
+    )
+
     return parser
 
 
@@ -158,6 +166,11 @@ def compute_fbank_kespeech_splits(args):
             keep_overlapping=False, min_duration=None
         )
 
+        if args.speed_perturb:
+            cut_set = (
+                    cut_set + cut_set.perturb_speed(0.9) + cut_set.perturb_speed(1.1)
+            )
+
         logging.info("Computing features")
         cut_set = cut_set.compute_and_store_features_batch(
             extractor=extractor,