From 6427c834dfd97b1f05c6659cdc7ccf010bf82fe1 Mon Sep 17 00:00:00 2001
From: 嘉渊 <wangjiaming.wjm@alibaba-inc.com>
Date: 星期一, 24 四月 2023 19:50:07 +0800
Subject: [PATCH] update
---
funasr/runtime/onnxruntime/src/Audio.cpp | 227 ++++++++++++++++++++++++++++++++++++++++++++++++++------
1 files changed, 202 insertions(+), 25 deletions(-)
diff --git a/funasr/runtime/onnxruntime/src/Audio.cpp b/funasr/runtime/onnxruntime/src/Audio.cpp
index f515a6d..bce3a90 100644
--- a/funasr/runtime/onnxruntime/src/Audio.cpp
+++ b/funasr/runtime/onnxruntime/src/Audio.cpp
@@ -3,7 +3,6 @@
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include <webrtc_vad.h>
#include "Audio.h"
@@ -25,8 +24,7 @@
out_idx = 1;
sum = 0;
};
- ~AudioWindow()
- {
+ ~AudioWindow(){
free(window);
};
int put(int val)
@@ -102,6 +100,11 @@
{
if (speech_buff != NULL) {
free(speech_buff);
+
+ }
+
+ if (speech_data != NULL) {
+
free(speech_data);
}
}
@@ -112,12 +115,20 @@
speech_len);
}
+float Audio::get_time_len()
+{
+ return (float)speech_len / 16000;
+ //speech_len);
+}
+
bool Audio::loadwav(const char *filename)
{
+ if (speech_data != NULL) {
+ free(speech_data);
+ }
if (speech_buff != NULL) {
free(speech_buff);
- free(speech_data);
}
offset = 0;
@@ -126,34 +137,198 @@
fp = fopen(filename, "rb");
if (fp == nullptr)
return false;
- fseek(fp, 0, SEEK_END);
- uint32_t nFileLen = ftell(fp);
- fseek(fp, 44, SEEK_SET);
+ fseek(fp, 0, SEEK_END); /*瀹氫綅鍒版枃浠舵湯灏�*/
+ uint32_t nFileLen = ftell(fp); /*寰楀埌鏂囦欢澶у皬*/
+ fseek(fp, 44, SEEK_SET); /*璺宠繃wav鏂囦欢澶�*/
speech_len = (nFileLen - 44) / 2;
speech_align_len = (int)(ceil((float)speech_len / align_size) * align_size);
speech_buff = (int16_t *)malloc(sizeof(int16_t) * speech_align_len);
- memset(speech_buff, 0, sizeof(int16_t) * speech_align_len);
- int ret = fread(speech_buff, sizeof(int16_t), speech_len, fp);
- fclose(fp);
- speech_data = (float *)malloc(sizeof(float) * speech_align_len);
- memset(speech_data, 0, sizeof(float) * speech_align_len);
- int i;
- float scale = 1;
+ if (speech_buff)
+ {
+ memset(speech_buff, 0, sizeof(int16_t) * speech_align_len);
+ int ret = fread(speech_buff, sizeof(int16_t), speech_len, fp);
+ fclose(fp);
- if (data_type == 1) {
- scale = 32768;
+ speech_data = (float*)malloc(sizeof(float) * speech_align_len);
+ memset(speech_data, 0, sizeof(float) * speech_align_len);
+ int i;
+ float scale = 1;
+
+ if (data_type == 1) {
+ scale = 32768;
+ }
+
+ for (i = 0; i < speech_len; i++) {
+ speech_data[i] = (float)speech_buff[i] / scale;
+ }
+
+ AudioFrame* frame = new AudioFrame(speech_len);
+ frame_queue.push(frame);
+
+
+ return true;
}
-
- for (i = 0; i < speech_len; i++) {
- speech_data[i] = (float)speech_buff[i] / scale;
- }
-
- AudioFrame *frame = new AudioFrame(speech_len);
- frame_queue.push(frame);
- return true;
+ else
+ return false;
}
+
+
+bool Audio::loadwav(const char* buf, int nFileLen)
+{
+
+
+
+ if (speech_data != NULL) {
+ free(speech_data);
+ }
+ if (speech_buff != NULL) {
+ free(speech_buff);
+ }
+
+ offset = 0;
+
+ size_t nOffset = 0;
+
+#define WAV_HEADER_SIZE 44
+
+ speech_len = (nFileLen - WAV_HEADER_SIZE) / 2;
+ speech_align_len = (int)(ceil((float)speech_len / align_size) * align_size);
+ speech_buff = (int16_t*)malloc(sizeof(int16_t) * speech_align_len);
+ if (speech_buff)
+ {
+ memset(speech_buff, 0, sizeof(int16_t) * speech_align_len);
+ memcpy((void*)speech_buff, (const void*)(buf + WAV_HEADER_SIZE), speech_len * sizeof(int16_t));
+
+
+ speech_data = (float*)malloc(sizeof(float) * speech_align_len);
+ memset(speech_data, 0, sizeof(float) * speech_align_len);
+ int i;
+ float scale = 1;
+
+ if (data_type == 1) {
+ scale = 32768;
+ }
+
+ for (i = 0; i < speech_len; i++) {
+ speech_data[i] = (float)speech_buff[i] / scale;
+ }
+
+
+ return true;
+ }
+ else
+ return false;
+
+}
+
+
+bool Audio::loadpcmwav(const char* buf, int nBufLen)
+{
+ if (speech_data != NULL) {
+ free(speech_data);
+ }
+ if (speech_buff != NULL) {
+ free(speech_buff);
+ }
+ offset = 0;
+
+ size_t nOffset = 0;
+
+
+
+ speech_len = nBufLen / 2;
+ speech_align_len = (int)(ceil((float)speech_len / align_size) * align_size);
+ speech_buff = (int16_t*)malloc(sizeof(int16_t) * speech_align_len);
+ if (speech_buff)
+ {
+ memset(speech_buff, 0, sizeof(int16_t) * speech_align_len);
+ memcpy((void*)speech_buff, (const void*)buf, speech_len * sizeof(int16_t));
+
+
+ speech_data = (float*)malloc(sizeof(float) * speech_align_len);
+ memset(speech_data, 0, sizeof(float) * speech_align_len);
+
+
+ int i;
+ float scale = 1;
+
+ if (data_type == 1) {
+ scale = 32768;
+ }
+
+ for (i = 0; i < speech_len; i++) {
+ speech_data[i] = (float)speech_buff[i] / scale;
+ }
+
+ AudioFrame* frame = new AudioFrame(speech_len);
+ frame_queue.push(frame);
+ return true;
+
+ }
+ else
+ return false;
+
+
+}
+
+bool Audio::loadpcmwav(const char* filename)
+{
+
+ if (speech_data != NULL) {
+ free(speech_data);
+ }
+ if (speech_buff != NULL) {
+ free(speech_buff);
+ }
+ offset = 0;
+
+ FILE* fp;
+ fp = fopen(filename, "rb");
+ if (fp == nullptr)
+ return false;
+ fseek(fp, 0, SEEK_END);
+ uint32_t nFileLen = ftell(fp);
+ fseek(fp, 0, SEEK_SET);
+
+ speech_len = (nFileLen) / 2;
+ speech_align_len = (int)(ceil((float)speech_len / align_size) * align_size);
+ speech_buff = (int16_t*)malloc(sizeof(int16_t) * speech_align_len);
+ if (speech_buff)
+ {
+ memset(speech_buff, 0, sizeof(int16_t) * speech_align_len);
+ int ret = fread(speech_buff, sizeof(int16_t), speech_len, fp);
+ fclose(fp);
+
+ speech_data = (float*)malloc(sizeof(float) * speech_align_len);
+ memset(speech_data, 0, sizeof(float) * speech_align_len);
+
+
+
+ int i;
+ float scale = 1;
+
+ if (data_type == 1) {
+ scale = 32768;
+ }
+
+ for (i = 0; i < speech_len; i++) {
+ speech_data[i] = (float)speech_buff[i] / scale;
+ }
+
+
+ AudioFrame* frame = new AudioFrame(speech_len);
+ frame_queue.push(frame);
+
+
+ return true;
+ }
+ else
+ return false;
+
+}
+
int Audio::fetch_chunck(float *&dout, int len)
{
@@ -163,7 +338,7 @@
} else if (offset == speech_align_len - len) {
dout = speech_data + offset;
offset = speech_align_len;
- // 涓存椂瑙ｅ喅
+ // 涓存椂瑙ｅ喅
AudioFrame *frame = frame_queue.front();
frame_queue.pop();
delete frame;
@@ -238,6 +413,7 @@
#define SPEECH_LEN_20S (16000 * 20)
#define SPEECH_LEN_30S (16000 * 30)
+/*
void Audio::split()
{
VadInst *handle = WebRtcVad_Create();
@@ -296,3 +472,4 @@
}
WebRtcVad_Free(handle);
}
+*/
\ No newline at end of file
--
Gitblit v1.9.1