From 3372b13d24aceef7002cfa0fc8222b3085c15110 Mon Sep 17 00:00:00 2001
From: 雾聪 <wucong.lyb@alibaba-inc.com>
Date: 星期五, 02 六月 2023 22:02:31 +0800
Subject: [PATCH] add fsmn-vad-online
---
funasr/runtime/onnxruntime/include/audio.h | 13 +++++++++----
1 files changed, 9 insertions(+), 4 deletions(-)
diff --git a/funasr/runtime/onnxruntime/include/audio.h b/funasr/runtime/onnxruntime/include/audio.h
index 1eabd3e..d2100a4 100644
--- a/funasr/runtime/onnxruntime/include/audio.h
+++ b/funasr/runtime/onnxruntime/include/audio.h
@@ -33,8 +33,9 @@
class Audio {
private:
- float *speech_data;
- int16_t *speech_buff;
+ float *speech_data=nullptr;
+ int16_t *speech_buff=nullptr;
+ char* speech_char=nullptr;
int speech_len;
int speech_align_len;
int offset;
@@ -47,18 +48,22 @@
Audio(int data_type, int size);
~Audio();
void Disp();
- bool LoadWav(const char* filename, int32_t* sampling_rate);
void WavResample(int32_t sampling_rate, const float *waveform, int32_t n);
bool LoadWav(const char* buf, int n_len, int32_t* sampling_rate);
+ bool LoadWav(const char* filename, int32_t* sampling_rate);
+ bool LoadWav2Char(const char* filename, int32_t* sampling_rate);
bool LoadPcmwav(const char* buf, int n_file_len, int32_t* sampling_rate);
bool LoadPcmwav(const char* filename, int32_t* sampling_rate);
+ bool LoadPcmwav2Char(const char* filename, int32_t* sampling_rate);
int FetchChunck(float *&dout, int len);
int Fetch(float *&dout, int &len, int &flag);
void Padding();
void Split(OfflineStream* offline_streamj);
- void Split(VadModel* vad_obj, vector<std::vector<int>>& vad_segments);
+ void Split(VadModel* vad_obj, vector<std::vector<int>>& vad_segments, bool input_finished=true);
float GetTimeLen();
int GetQueueSize() { return (int)frame_queue.size(); }
+ char* GetSpeechChar(){return speech_char;}
+ int GetSpeechLen(){return speech_len;}
};
} // namespace funasr
--
Gitblit v1.9.1