Use high_freq -400 in computing fbank features. (#528)

See also k2-fsa/sherpa-onnx#514
k2-fsa · Jan 4, 2024 · 62f9697 · 62f9697
1 parent 30ed540
commit 62f9697
Show file tree

Hide file tree

Showing 8 changed files with 23 additions and 1 deletion.
diff --git a/sherpa/bin/offline_ctc_asr.py b/sherpa/bin/offline_ctc_asr.py
@@ -303,6 +303,7 @@ def create_recognizer(args):
 
     feat_config.fbank_opts.frame_opts.samp_freq = 16000
     feat_config.fbank_opts.mel_opts.num_bins = 80
+    feat_config.fbank_opts.mel_opts.high_freq = -400
     feat_config.fbank_opts.frame_opts.dither = 0
 
     feat_config.normalize_samples = args.normalize_samples
@@ -377,7 +378,9 @@ def main():
 
 if __name__ == "__main__":
     torch.manual_seed(20230104)
-    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"  # noqa
+    formatter = (
+        "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"  # noqa
+    )
     logging.basicConfig(format=formatter, level=logging.INFO)
 
     main()
diff --git a/sherpa/bin/offline_ctc_server.py b/sherpa/bin/offline_ctc_server.py
@@ -302,6 +302,7 @@ def create_recognizer(args) -> sherpa.OfflineRecognizer:
 
     feat_config.fbank_opts.frame_opts.samp_freq = args.sample_rate
     feat_config.fbank_opts.mel_opts.num_bins = args.feat_dim
+    feat_config.fbank_opts.mel_opts.high_freq = -400
     feat_config.fbank_opts.frame_opts.dither = 0
 
     feat_config.normalize_samples = args.normalize_samples

diff --git a/sherpa/bin/offline_transducer_asr.py b/sherpa/bin/offline_transducer_asr.py
@@ -403,6 +403,7 @@ def create_recognizer(args) -> sherpa.OfflineRecognizer:
 
     feat_config.fbank_opts.frame_opts.samp_freq = args.sample_rate
     feat_config.fbank_opts.mel_opts.num_bins = args.feat_dim
+    feat_config.fbank_opts.mel_opts.high_freq = -400
     feat_config.fbank_opts.frame_opts.dither = 0
 
     fast_beam_search_config = sherpa.FastBeamSearchConfig(

diff --git a/sherpa/bin/offline_transducer_server.py b/sherpa/bin/offline_transducer_server.py
@@ -636,6 +636,7 @@ def create_recognizer(args) -> sherpa.OfflineRecognizer:
 
     feat_config.fbank_opts.frame_opts.samp_freq = args.sample_rate
     feat_config.fbank_opts.mel_opts.num_bins = args.feat_dim
+    feat_config.fbank_opts.mel_opts.high_freq = -400
     feat_config.fbank_opts.frame_opts.dither = 0
 
     fast_beam_search_config = sherpa.FastBeamSearchConfig(

diff --git a/sherpa/bin/online_transducer_asr.py b/sherpa/bin/online_transducer_asr.py
@@ -392,6 +392,7 @@ def create_recognizer(args) -> sherpa.OnlineRecognizer:
 
     feat_config.fbank_opts.frame_opts.samp_freq = args.sample_rate
     feat_config.fbank_opts.mel_opts.num_bins = args.feat_dim
+    feat_config.fbank_opts.mel_opts.high_freq = -400
     feat_config.fbank_opts.frame_opts.dither = 0
 
     fast_beam_search_config = sherpa.FastBeamSearchConfig(

diff --git a/sherpa/bin/streaming_server.py b/sherpa/bin/streaming_server.py
@@ -400,6 +400,7 @@ def create_recognizer(args) -> sherpa.OnlineRecognizer:
 
     feat_config.fbank_opts.frame_opts.samp_freq = args.sample_rate
     feat_config.fbank_opts.mel_opts.num_bins = args.feat_dim
+    feat_config.fbank_opts.mel_opts.high_freq = -400
     feat_config.fbank_opts.frame_opts.dither = 0
 
     fast_beam_search_config = sherpa.FastBeamSearchConfig(

diff --git a/sherpa/python/test/test_offline_recognizer.py b/sherpa/python/test/test_offline_recognizer.py
@@ -34,6 +34,7 @@ def test_icefall_ctc_model(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = 16000
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         config = sherpa.OfflineRecognizerConfig(
@@ -72,6 +73,7 @@ def test_icefall_ctc_model_hlg_decoding(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = 16000
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         ctc_decoder_config = sherpa.OfflineCtcDecoderConfig(hlg=hlg)
@@ -112,6 +114,7 @@ def test_wenet_ctc_model(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = 16000
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         config = sherpa.OfflineRecognizerConfig(
@@ -185,6 +188,7 @@ def test_nemo_ctc_en_model(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = 16000
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         # The following option is very important.
@@ -236,6 +240,7 @@ def test_nemo_ctc_zh_model(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = 16000
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         # The following option is very important.
@@ -288,6 +293,7 @@ def test_icefall_transducer_model(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = 16000
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         config = sherpa.OfflineRecognizerConfig(

diff --git a/sherpa/python/test/test_online_recognizer.py b/sherpa/python/test/test_online_recognizer.py
@@ -82,6 +82,7 @@ def test_icefall_asr_librispeech_conv_emformer_transducer_stateless2_2022_07_05(
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -190,6 +191,7 @@ def test_icefall_asr_wenetspeech_pruned_transducer_stateless5_streaming(
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -316,6 +318,7 @@ def test_icefall_asr_conv_emformer_transducer_stateless2_zh(
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -398,6 +401,7 @@ def test_icefall_librispeech_streaming_pruned_transducer_stateless4_20220625(
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -521,6 +525,7 @@ def test_cefall_asr_librispeech_lstm_transducer_stateless2_2022_09_03(
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -641,6 +646,7 @@ def test_icefall_asr_librispeech_pruned_stateless_emformer_rnnt2_2022_06_01(
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -744,6 +750,7 @@ def test_k2fsa_zipformer_chinese_english_mixed(self):
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")
@@ -829,6 +836,7 @@ def test_icefall_asr_librispeech_pruned_transducer_stateless7_streaming_2022_12_
 
         feat_config.fbank_opts.frame_opts.samp_freq = expected_sample_rate
         feat_config.fbank_opts.mel_opts.num_bins = 80
+        feat_config.fbank_opts.mel_opts.high_freq = -400
         feat_config.fbank_opts.frame_opts.dither = 0
 
         print("--------------------greedy search--------------------")