Merge pull request #263 from alibaba-damo-academy/tmp

update paraformer streaming recipe

egs_modelscope/asr/paraformer/speech_paraformer_asr_nat-zh-cn-16k-common-vocab8404-online/infer.py

@@ -13,7 +13,7 @@ inference_pipeline = pipeline(
     model='damo/speech_paraformer_asr_nat-zh-cn-16k-common-vocab8404-online',
     model_revision='v1.0.2')
 
-waveform, sample_rate = torchaudio.load("asr_example_zh.wav")
+waveform, sample_rate = torchaudio.load("waihu.wav")
 speech_length = waveform.shape[1]
 speech = waveform[0]
 
@@ -28,7 +28,7 @@ speech_buffer = speech
 speech_cache = []
 final_result = ""
 
-while len(speech_buffer) > 0:
+while len(speech_buffer) >= 960:
     if first_chunk:
         if len(speech_buffer) >= 14400:
             rec_result = inference_pipeline(audio_in=speech_buffer[0:14400], param_dict=param_dict)
@@ -47,11 +47,11 @@ while len(speech_buffer) > 0:
             rec_result = inference_pipeline(audio_in=speech_buffer[:19200], param_dict=param_dict)
             speech_buffer = speech_buffer[9600:]
         else:
-            cache_en["stride"] = len(speech_buffer) // 960
+            cache_en["stride"] = len(speech_buffer) // 960 
             cache_en["pad_right"] = 0
             rec_result = inference_pipeline(audio_in=speech_buffer, param_dict=param_dict)
             speech_buffer = []
-    if rec_result['text'] != "sil":
+    if len(rec_result) !=0 and rec_result['text'] != "sil":
         final_result += rec_result['text']
     print(rec_result)
 print(final_result)