1 #ifndef _RKAUDIO_PREPROCESS_H_
2 #define _RKAUDIO_PREPROCESS_H_
3
4 #include <stdio.h>
5 #include <stdlib.h>
6 #ifdef __cplusplus
7 extern "C" {
8 #endif
9 #define NUM_CHANNEL 8
10 #define NUM_REF_CHANNEL 1
11 #define NUM_DROP_CHANNEL 0
12 #define REF_POSITION 1
13 //static short int Array[NUM_SRC_CHANNEL] = { 9,7,5,3,2,4,6,8 };
14 //static short int Array[NUM_CHANNEL] = {2, 3, 0, 1}; //src first, ref second
15 static short int Array[NUM_CHANNEL] = { 1, 0, 3, 2, 5, 4, 6, 7 };
16 //Array[NUM_SRC_CHANNEL] = { 2,3,4,5,6,7};
17 /**********************EQ Parameter**********************/
18 static short EqPara_16k[5][13] =
19 {
20 //filter_bank 1
21 {-1 ,-1 ,-1 ,-1 ,-2 ,-1 ,-1 ,-1 ,-1 ,-1 ,-1 ,-2 ,-3 },
22 //filter_bank 2
23 {-1 ,-1 ,-1 ,-1 ,-2 ,-2 ,-3 ,-5 ,-3 ,-2 ,-1 ,-1 ,-2 },
24 //filter_bank 3
25 {-2 ,-5 ,-9 ,-4 ,-2 ,-2 ,-1 ,-5 ,-5 ,-11 ,-20 ,-11 ,-5 },
26 //filter_bank 4
27 {-5 ,-1 ,-7 ,-7 ,-19 ,-40 ,-20 ,-9 ,-10 ,-1 ,-20 ,-24 ,-60 },
28 //filter_bank 5
29 {-128 ,-76 ,-40 ,-44 ,-1 ,-82 ,-111 ,-383 ,-1161 ,-1040 ,-989 ,-3811 ,32764 },
30 };
31
32 /**********************NLP Parameter**********************/
33 static short int ashwAecBandNlpPara_16k[8][2] =
34 {
35 /* BandPassThd SuperEstFactor*/
36 { 10, 6 }, /* Hz: 0 - 300 */
37 { 10, 10 }, /* Hz: 300 - 575 */
38 { 10, 10 }, /* Hz: 575 - 950 */
39 { 5, 10 }, /* Hz: 950 - 1425 */
40 { 5, 10 }, /* Hz: 1425 - 2150 */
41 { 5, 10 }, /* Hz: 2150 - 3350 */
42 { 0, 6 }, /* Hz: 3350 - 5450 */
43 { 0, 6 }, /* Hz: 5450 - 8000 */
44 };
45 /*************************************************/
46 /*The Main Enable which used to control the AEC,BF and RX*/
47 typedef enum RKAUDIOEnable_
48 {
49 RKAUDIO_EN_AEC = 1 << 0,
50 RKAUDIO_EN_BF = 1 << 1,
51 RKAUDIO_EN_RX = 1 << 2,
52 RKAUDIO_EN_CMD = 1 << 3,
53 } RKAUDIOEnable;
54
55 /* The Sub-Enable which used to control the AEC,BF and RX*/
56 typedef enum RKAecEnable_
57 {
58 EN_DELAY = 1 << 0,
59 EN_ARRAY_RESET = 1<<1,
60 EN_HOWLING = 1 << 2,
61 } RKAecEnable;
62 typedef enum RKPreprocessEnable_
63 {
64 EN_Fastaec = 1 << 0,
65 EN_Wakeup = 1 << 1,
66 EN_Dereverberation = 1 << 2,
67 EN_Nlp = 1 << 3,
68 EN_AES = 1<<4,
69 EN_Agc = 1 << 5,
70 EN_Anr = 1 << 6,
71 EN_GSC = 1 << 7,
72 GSC_Method = 1 << 8,
73 EN_Fix = 1 << 9,
74 EN_STDT = 1 << 10,
75 EN_CNG = 1<<11,
76 EN_EQ = 1<<12,
77 EN_CHN_SELECT = 1<<13,
78 } RKPreprocessEnable;
79 typedef enum RkaudioRxEnable_
80 {
81 EN_RX_Anr = 1 << 0,
82 } RkaudioRxEnable;
83 /*****************************************/
84
85 /* Set the three Main Para which used to initialize the AEC,BF and RX*/
86 typedef struct SKVAECParameter_ {
87 int pos;
88 int drop_ref_channel;
89 int model_aec_en;
90 int delay_len;
91 int look_ahead;
92 short int *Array_list;
93 } SKVAECParameter;
94 typedef struct SKVPreprocessParam_
95 {
96 /* Parameters of agc */
97 int model_bf_en;
98 int ref_pos;
99 int Targ;
100 int num_ref_channel;
101 int drop_ref_channel;
102 void* dereverb_para;
103 void* aes_para;
104 void* nlp_para;
105 void* anr_para;
106 void* agc_para;
107 void* cng_para;
108 void* dtd_para;
109 void* eq_para;
110 }SKVPreprocessParam;
111 typedef struct RkaudioRxParam_
112 {
113 /* Parameters of agc */
114 int model_rx_en;
115 void* anr_para;
116 }RkaudioRxParam;
117 /****************************************/
118 /*The param struct of sub-mudule of AEC,BF and RX*/
119 typedef struct RKAudioDereverbParam_
120 {
121 int rlsLg;
122 int curveLg;
123 int delay;
124 float forgetting;
125 float T60;
126 float coCoeff;
127 } RKAudioDereverbParam;
128
129 typedef struct RKAudioAESParameter_ {
130 float Beta_Up;
131 float Beta_Down;
132 } RKAudioAESParameter;
133
134 typedef struct SKVNLPParameter_ {
135 short int g_ashwAecBandNlpPara_16k[8][2];
136 } SKVNLPParameter;
137 typedef struct SKVANRParam_ {
138 float noiseFactor;
139 int swU;
140 float PsiMin;
141 float PsiMax;
142 float fGmin;
143 } SKVANRParam;
144
145 typedef struct RKAGCParam_ {
146 /* 新版AGC参数 */
147 float attack_time; /* 触发时间，即AGC增益下降所需要的时间 */
148 float release_time; /* 施放时间，即AGC增益上升所需要的时间 */
149 float max_gain; /* 最大增益，同时也是线性段增益，单位：dB */
150 float max_peak; /* 经AGC处理后，输出语音的最大能量，范围：单位：dB */
151 float fRth0; /* 扩张段结束能量dB阈值，同时也是线性段开始阈值 */
152 float fRk0; /* 扩张段斜率 */
153 float fRth1; /* 压缩段起始能量dB阈值，同时也是线性段结束阈值 */
154
155 /* 无效参数 */
156 int fs; /* 数据采样率 */
157 int frmlen; /* 处理帧长 */
158 float attenuate_time; /* 噪声衰减时间，即噪声段增益衰减到1所需的时间 */
159 float fRth2; /* 压缩段起始能量dB阈值 */
160 float fRk1; /* 扩张段斜率 */
161 float fRk2; /* 扩张段斜率 */
162 float fLineGainDb; /* 线性段提升dB数 */
163 int swSmL0; /* 扩张段时域平滑点数 */
164 int swSmL1; /* 线性段时域平滑点数 */
165 int swSmL2; /* 压缩段时域平滑点数 */
166
167 } RKAGCParam;
168
169 typedef struct RKCNGParam_
170 {
171 /*CNG Parameter*/
172 float fGain; /* INT16 Q0 施加舒适噪声幅度比例 */
173 float fMpy; /* INT16 Q0 白噪随机数生成幅度 */
174 float fSmoothAlpha; /* 舒适噪声平滑系数 */
175 float fSpeechGain; /* 根据语音能量额外施加舒适噪声比例增益 */
176 } RKCNGParam;
177
178 typedef struct RKDTDParam_
179 {
180 float ksiThd_high; /* 单双讲判决阈值 */
181 float ksiThd_low; /* 单双讲判决阈值 */
182
183 }RKDTDParam;
184
185 typedef struct RKaudioEqParam_ {
186 int shwParaLen; // 滤波器系数个数
187 short pfCoeff[5][13]; // 滤波器系数
188 } RKaudioEqParam;
189
190
191 /* Set the Sub-Para which used to initialize the Dereverb*/
rkaudio_dereverb_param_init()192 inline static void * rkaudio_dereverb_param_init(){
193 RKAudioDereverbParam* param = (RKAudioDereverbParam*)malloc(sizeof(RKAudioDereverbParam));
194 param->rlsLg = 4; /* RLS滤波器阶数 */
195 param->curveLg = 10; /* 分布曲线阶数 */
196 param->delay = 2; /* RLS滤波器延时 */
197 param->forgetting = 0.98; /* RLS滤波器遗忘因子 */
198 param->T60 = 0.5;//1.5; /* 混响时间估计值（单位：s），越大，去混响能力越强，但是越容易过消除 */
199 param->coCoeff = 1.0; /* 互相干性调整系数，防止过消除，越大能力越强，建议取值：0.5到2之间 */
200 return (void*)param;
201 }
rkaudio_aes_param_init()202 inline static void* rkaudio_aes_param_init() {
203 RKAudioAESParameter* param = (RKAudioAESParameter*)malloc(sizeof(RKAudioAESParameter));
204 //param->Beta_Up = 0.001f; /* 上升速度 */
205 //param->Beta_Down = 0.005f; /* 下降速度 */
206 param->Beta_Up = 0.005f; /* 上升速度 */
207 param->Beta_Down = 0.001f; /* 下降速度 */
208 //param->Beta_Up = 0.0005f; /* 上升速度 */
209 //param->Beta_Down = 0.01f; /* 下降速度 */
210 return (void*)param;
211 }
212 /* Set the Sub-Para which used to initialize the NLP*/
rkaudio_nlp_param_init()213 inline static void* rkaudio_nlp_param_init()
214 {
215 SKVNLPParameter* param = (SKVNLPParameter*)malloc(sizeof(SKVNLPParameter));
216 for (int i = 0; i < 8; i++)
217 {
218 for (int j = 0; j < 2; j++)
219 {
220 param->g_ashwAecBandNlpPara_16k[i][j] = ashwAecBandNlpPara_16k[i][j];
221 }
222 }
223 return (void*)param;
224 }
225 /* Set the Sub-Para which used to initialize the ANR*/
rkaudio_anr_param_init()226 inline static void *rkaudio_anr_param_init() {
227 SKVANRParam* param = (SKVANRParam*)malloc(sizeof(SKVANRParam));
228 /* anr parameters */
229 param->noiseFactor = 0.88;
230 param->swU = 10; //2
231 //param->PsiMin = 0.2; //0.05
232 //param->PsiMax = 0.316; //0.516
233 //param->fGmin = 0.1; //0.1
234 param->PsiMin = 0.01; //0.05
235 param->PsiMax = 0.516; //0.516
236 param->fGmin = 0.05; //0.1
237 return (void*)param;
238 }
239 /* Set the Sub-Para which used to initialize the AGC*/
rkaudio_agc_param_init()240 inline static void* rkaudio_agc_param_init()
241 {
242 RKAGCParam* param = (RKAGCParam*)malloc(sizeof(RKAGCParam));
243
244 /* 新版AGC参数 */
245 param->attack_time = 200.0; /* 触发时间，即AGC增益上升所需要的时间 */
246 param->release_time = 200.0; /* 施放时间，即AGC增益下降所需要的时间 */
247 //param->max_gain = 35.0; /* 最大增益，同时也是线性段增益，单位：dB */
248 param->max_gain = 30; /* 最大增益，同时也是线性段增益，单位：dB */
249 param->max_peak = -1.0; /* 经AGC处理后，输出语音的最大能量，范围：单位：dB */
250 param->fRk0 = 2; /* 扩张段斜率 */
251 param->fRth2 = -35; /* 压缩段起始能量dB阈值，同时也是线性段结束阈值，增益逐渐降低，注意 fRth2 + max_gain < max_peak */
252 param->fRth1 = -80; /* 扩张段结束能量dB阈值，同时也是线性段开始阈值，能量高于改区域以max_gain增益 */
253 param->fRth0 = -85; /* 噪声门阈值 */
254
255 /* 无效参数 */
256 param->fs = 16000; /* 数据采样率 */
257 param->frmlen = 256; /* 处理帧长 */
258 param->attenuate_time = 1000; /* 噪声衰减时间，即噪声段增益衰减到1所需的时间 */
259 param->fRk1 = 0.8; /* 扩张段斜率 */
260 param->fRk2 = 0.4; /* 扩张段斜率 */
261 param->fLineGainDb = -25.0f; /* 低于该值，起始的attenuate_time(ms)内不做增益 */
262 param->swSmL0 = 40; /* 扩张段时域平滑点数 */
263 param->swSmL1 = 80; /* 线性段时域平滑点数 */
264 param->swSmL2 = 80; /* 压缩段时域平滑点数 */
265
266 return (void*)param;
267 }
268 /* Set the Sub-Para which used to initialize the CNG*/
rkaudio_cng_param_init()269 inline static void* rkaudio_cng_param_init()
270 {
271 RKCNGParam* param = (RKCNGParam*)malloc(sizeof(RKCNGParam));
272 /* cng paremeters */
273 param->fSmoothAlpha = 0.99f; /* INT16 Q15 施加舒适噪声平滑度 */
274 param->fSpeechGain = 0; /* INT16 Q15 施加舒适噪声语音纹理模拟程度 */
275 param->fGain = 20.0; /* INT16 Q0 施加舒适噪声幅度比例 */
276 param->fMpy = 20; /* INT16 Q0 白噪随机数生成幅度 */
277 return (void*)param;
278 }
279 /* Set the Sub-Para which used to initialize the DTD*/
rkaudio_dtd_param_init()280 inline static void* rkaudio_dtd_param_init()
281 {
282 RKDTDParam* param = (RKDTDParam*)malloc(sizeof(RKDTDParam));
283 /* dtd paremeters*/
284 param->ksiThd_high = 0.60f; /* 单双讲判决阈值 */
285 param->ksiThd_low = 0.30f;
286 return (void*)param;
287 }
288 /* Set the Sub-Para which used to initialize the AEC*/
289
rkaudio_aec_param_init()290 inline static void* rkaudio_aec_param_init()
291 {
292 SKVAECParameter* param = (SKVAECParameter*)malloc(sizeof(SKVAECParameter));
293 param->pos = REF_POSITION;
294 param->drop_ref_channel = NUM_DROP_CHANNEL;
295 param->model_aec_en = 0;//param->model_aec_en = EN_DELAY;
296 param->delay_len = 0;
297 param->look_ahead = 0;
298 param->Array_list = Array;
299 return (void*)param;
300 }
301
rkaudio_eq_param_init()302 inline static void * rkaudio_eq_param_init(){
303 RKaudioEqParam* param = (RKaudioEqParam*)malloc(sizeof(RKaudioEqParam));
304 param->shwParaLen = 65;
305 int i, j;
306 for (i = 0; i < 5; i++){
307 for (j = 0; j < 13; j++) {
308 param->pfCoeff[i][j] = EqPara_16k[i][j];
309 }
310 }
311 return (void*)param;
312 }
313
314 /* Set the Sub-Para which used to initialize the BF*/
rkaudio_preprocess_param_init()315 inline static void* rkaudio_preprocess_param_init()
316 {
317 SKVPreprocessParam* param = (SKVPreprocessParam*)malloc(sizeof(SKVPreprocessParam));
318 //param->model_bf_en = EN_Fastaec |EN_AES |EN_Wakeup;
319 //param->model_bf_en = EN_Agc | EN_Anr | EN_Dereverberation;
320 param->model_bf_en = EN_Fastaec | EN_AES | EN_Anr | EN_Agc | EN_GSC | EN_Dereverberation;
321 //param->model_bf_en = EN_Wakeup ;
322 //param->model_bf_en = EN_Fix;
323 //param->model_bf_en = EN_Anr | EN_GSC |EN_Agc;
324 //param->model_bf_en = EN_Fastaec | EN_AES | EN_Anr | EN_Agc;
325 param->Targ = 4;
326 param->ref_pos = REF_POSITION;
327 param->num_ref_channel = NUM_REF_CHANNEL;
328 param->drop_ref_channel = NUM_DROP_CHANNEL;
329 param->dereverb_para = rkaudio_dereverb_param_init();
330 param->aes_para = rkaudio_aes_param_init();
331 param->nlp_para = rkaudio_nlp_param_init();
332 param->anr_para = rkaudio_anr_param_init();
333 param->agc_para = rkaudio_agc_param_init();
334 param->cng_para = rkaudio_cng_param_init();
335 param->dtd_para = rkaudio_dtd_param_init();
336 param->eq_para = rkaudio_eq_param_init();
337 return (void*)param;
338 }
339 /* Set the Sub-Para which used to initialize the RX*/
rkaudio_rx_param_init()340 inline static void* rkaudio_rx_param_init()
341 {
342 RkaudioRxParam* param = (RkaudioRxParam*)malloc(sizeof(RkaudioRxParam));
343 param->anr_para = rkaudio_anr_param_init();
344 param->model_rx_en = EN_RX_Anr;
345 return (void*)param;
346 }
347 typedef struct RKAUDIOParam_
348 {
349 int model_en;
350 void *aec_param;
351 void *bf_param;
352 void* rx_param;
353 } RKAUDIOParam;
354
rkaudio_aec_param_destory(void * param_)355 inline static void rkaudio_aec_param_destory(void* param_)
356 {
357 SKVAECParameter* param = (SKVAECParameter*)param_;
358 free(param); param = NULL;
359 }
360
rkaudio_preprocess_param_destory(void * param_)361 inline static void rkaudio_preprocess_param_destory(void* param_)
362 {
363 SKVPreprocessParam* param = (SKVPreprocessParam*)param_;
364 free(param->dereverb_para); param->dereverb_para = NULL;
365 free(param->aes_para); param->aes_para = NULL;
366 free(param->anr_para); param->anr_para = NULL;
367 free(param->agc_para); param->agc_para = NULL;
368 free(param->nlp_para); param->nlp_para = NULL;
369 free(param->cng_para); param->cng_para = NULL;
370 free(param->dtd_para); param->dtd_para = NULL;
371 free(param->eq_para); param->eq_para = NULL;
372 free(param); param = NULL;
373 }
374
rkaudio_rx_param_destory(void * param_)375 inline static void rkaudio_rx_param_destory(void* param_)
376 {
377 RkaudioRxParam* param = (RkaudioRxParam*)param_;
378 free(param->anr_para); param->anr_para = NULL;
379 free(param); param = NULL;
380 }
381
rkaudio_param_deinit(void * param_)382 inline static void rkaudio_param_deinit(void* param_)
383 {
384 RKAUDIOParam* param = (RKAUDIOParam*)param_;
385 rkaudio_aec_param_destory(param->aec_param);
386 rkaudio_preprocess_param_destory(param->bf_param);
387 rkaudio_rx_param_destory(param->rx_param);
388 }
389
390 void *rkaudio_preprocess_init(int rate, int bits, int src_chan, int ref_chan, RKAUDIOParam *param);
391 void rkaudio_param_printf(int src_chan, int ref_chan, RKAUDIOParam* param);
392 int rkaudio_Doa_invoke(void* st_ptr);
393 int rkaudio_Cir_Doa_invoke(void* st_ptr, int* ang_doa, int* pth_doa);
394 int rkaudio_preprocess_get_cmd_id(void* st_ptr, float* asr_score, int* cmd_id);
395
396 void rkaudio_preprocess_destory(void *st_ptr);
397
398 int rkaudio_preprocess_short(void *st_ptr, short *in, short *out, int in_size, int *is_wakeup);
399 int rkaudio_rx_short(void* st_ptr, short* in, short* out);
400 int rkaudio_preprocess_key_short(void *st_ptr, short *in, short *out, int in_size);
401 int rkaudio_preprocess_command_short(void *st_ptr, short *in, int in_size, float *asr_scor);
402 void rkaudio_asr_set_param(float min, float max, float keep);
403
404 void rkaudio_param_deinit(void* param_);
405 #ifdef __cplusplus
406 }
407 #endif
408
409 #endif // _RKAUDIO_PREPROCESS_H_
410