| | |
|---|
| | | 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] |
|---|
| | | |
|---|
| | |
|---|
| | | 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) |
|---|
| | |
|---|
| | | 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) |
|---|
