1+ #pragma once
2+ #include "stdint.h"
3+ #include "esp_wn_iface.h"
4+ #include "esp_wn_models.h"
5+ #include "esp_vad.h"
6+
7+ //AFE: Audio Front-End
8+ //SR: Speech Recognition
9+ //afe_sr/AFE_SR: the audio front-end for speech recognition
10+
11+ //Set AFE_SR mode
12+ typedef enum {
13+ SR_MODE_LOW_COST = 0 ,
14+ SR_MODE_HIGH_PERF = 1
15+ } afe_sr_mode_t ;
16+
17+ typedef enum {
18+ AFE_MEMORY_ALLOC_MORE_INTERNAL = 1 , // malloc with more internal ram
19+ AFE_MEMORY_ALLOC_INTERNAL_PSRAM_BALANCE = 2 , // malloc with internal ram and psram in balance
20+ AFE_MEMORY_ALLOC_MORE_PSRAM = 3 // malloc with more psram
21+ } afe_memory_alloc_mode_t ;
22+
23+ typedef enum {
24+ AFE_MN_PEAK_AGC_MODE_1 = -5 , // The peak amplitude of audio fed to multinet is -5dB
25+ AFE_MN_PEAK_AGC_MODE_2 = -4 , // The peak amplitude of audio fed to multinet is -4dB
26+ AFE_MN_PEAK_AGC_MODE_3 = -3 , // The peak amplitude of audio fed to multinet is -3dB
27+ AFE_MN_PEAK_NO_AGC = 0 , // There is no agc gain
28+ } afe_mn_peak_agc_mode_t ;
29+
30+ typedef struct {
31+ int total_ch_num ; // total channel num. It must be: total_ch_num = mic_num + ref_num
32+ int mic_num ; // mic channel num
33+ int ref_num ; // reference channel num
34+ } afe_pcm_config_t ;
35+
36+ typedef struct {
37+ bool aec_init ;
38+ bool se_init ;
39+ bool vad_init ;
40+ bool wakenet_init ;
41+ bool voice_communication_init ;
42+ bool voice_communication_agc_init ; // AGC swich for voice communication
43+ int voice_communication_agc_gain ; // AGC gain(dB) for voice communication
44+ vad_mode_t vad_mode ; // The value can be: VAD_MODE_0, VAD_MODE_1, VAD_MODE_2, VAD_MODE_3, VAD_MODE_4
45+ char * wakenet_model_name ; // The model name of wakenet
46+ det_mode_t wakenet_mode ;
47+ afe_sr_mode_t afe_mode ;
48+ int afe_perferred_core ;
49+ int afe_perferred_priority ;
50+ int afe_ringbuf_size ;
51+ afe_memory_alloc_mode_t memory_alloc_mode ;
52+ afe_mn_peak_agc_mode_t agc_mode ; // The agc mode for ASR
53+ afe_pcm_config_t pcm_config ; // Config the channel num of original data which is fed to the afe feed function.
54+ } afe_config_t ;
55+
56+
57+ #if CONFIG_IDF_TARGET_ESP32
58+ #define AFE_CONFIG_DEFAULT () { \
59+ .aec_init = true, \
60+ .se_init = true, \
61+ .vad_init = true, \
62+ .wakenet_init = true, \
63+ .voice_communication_init = false, \
64+ .voice_communication_agc_init = false, \
65+ .voice_communication_agc_gain = 15, \
66+ .vad_mode = VAD_MODE_3, \
67+ .wakenet_model_name = NULL, \
68+ .wakenet_mode = DET_MODE_90, \
69+ .afe_mode = SR_MODE_HIGH_PERF, \
70+ .afe_perferred_core = 0, \
71+ .afe_perferred_priority = 5, \
72+ .afe_ringbuf_size = 50, \
73+ .memory_alloc_mode = AFE_MEMORY_ALLOC_INTERNAL_PSRAM_BALANCE, \
74+ .agc_mode = AFE_MN_PEAK_AGC_MODE_2, \
75+ .pcm_config.total_ch_num = 2, \
76+ .pcm_config.mic_num = 1, \
77+ .pcm_config.ref_num = 1, \
78+ }
79+ #elif CONFIG_IDF_TARGET_ESP32S3
80+ #define AFE_CONFIG_DEFAULT () { \
81+ .aec_init = true, \
82+ .se_init = true, \
83+ .vad_init = true, \
84+ .wakenet_init = true, \
85+ .voice_communication_init = false, \
86+ .voice_communication_agc_init = false, \
87+ .voice_communication_agc_gain = 15, \
88+ .vad_mode = VAD_MODE_3, \
89+ .wakenet_model_name = NULL, \
90+ .wakenet_mode = DET_MODE_2CH_90, \
91+ .afe_mode = SR_MODE_LOW_COST, \
92+ .afe_perferred_core = 0, \
93+ .afe_perferred_priority = 5, \
94+ .afe_ringbuf_size = 50, \
95+ .memory_alloc_mode = AFE_MEMORY_ALLOC_MORE_PSRAM, \
96+ .agc_mode = AFE_MN_PEAK_AGC_MODE_2, \
97+ .pcm_config.total_ch_num = 3, \
98+ .pcm_config.mic_num = 2, \
99+ .pcm_config.ref_num = 1, \
100+ }
101+ #endif
0 commit comments