airosss commited on
Commit
ad11b49
·
verified ·
1 Parent(s): c80ef23

Update analysis_schema.json

Browse files
Files changed (1) hide show
  1. analysis_schema.json +135 -70
analysis_schema.json CHANGED
@@ -1,79 +1,144 @@
1
  {
2
  "$schema": "https://json-schema.org/draft/2020-12/schema",
3
- "$id": "https://aiangel/schema/voice-analysis/1.0",
4
- "title": "AI Angel Voice Analysis v1.0",
5
- "type": "object",
6
- "required": ["version", "aggregates", "words", "transcript"],
7
- "properties": {
8
- "version": {
9
- "type": "string",
10
- "const": "1.0",
11
- "description": "Версия формата файла."
12
- },
13
- "aggregates": {
14
  "type": "object",
15
- "required": [
16
- "duration_sec",
17
- "voiced_duration_sec",
18
- "voiced_ratio",
19
- "rms_dbfs_mean",
20
- "f0_mean_hz",
21
- "f0_median_hz",
22
- "f0_std_hz",
23
- "f0_stability",
24
- "words_count",
25
- "speech_rate_wps",
26
- "vad"
27
- ],
28
  "properties": {
29
- "duration_sec": { "type": "number", "minimum": 0 },
30
- "voiced_duration_sec": { "type": "number", "minimum": 0 },
31
- "voiced_ratio": { "type": "number", "minimum": 0, "maximum": 1 },
32
- "rms_dbfs_mean": { "type": ["number", "null"] },
33
- "f0_mean_hz": { "type": ["number", "null"], "minimum": 0 },
34
- "f0_median_hz": { "type": ["number", "null"], "minimum": 0 },
35
- "f0_std_hz": { "type": ["number", "null"], "minimum": 0 },
36
- "f0_stability": { "type": ["number", "null"], "minimum": 0, "maximum": 1 },
37
- "words_count": { "type": "integer", "minimum": 0 },
38
- "speech_rate_wps": { "type": "number", "minimum": 0 },
39
- "vad": { "type": "string" }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
40
  },
41
  "additionalProperties": false
42
  },
43
- "words": {
44
- "type": "array",
45
- "description": "Таблица по словам.",
46
- "items": {
47
- "type": "object",
48
- "required": [
49
- "start_s", "end_s", "duration_s", "pre_pause_ms", "word",
50
- "prob", "f0_mean_hz", "f0_z", "rms_dbfs", "rms_z",
51
- "accent_score", "accent_flag", "speed_local_sylps", "duration_z"
52
- ],
53
- "properties": {
54
- "start_s": { "type": "number", "minimum": 0 },
55
- "end_s": { "type": "number", "minimum": 0 },
56
- "duration_s": { "type": "number", "minimum": 0 },
57
- "pre_pause_ms": { "type": "integer", "minimum": 0 },
58
- "word": { "type": "string" },
59
- "prob": { "type": ["number", "null"], "minimum": 0, "maximum": 1 },
60
- "f0_mean_hz": { "type": ["number", "null"], "minimum": 0 },
61
- "f0_z": { "type": ["number", "null"] },
62
- "rms_dbfs": { "type": ["number", "null"] },
63
- "rms_z": { "type": ["number", "null"] },
64
- "accent_score": { "type": "number", "minimum": 0, "maximum": 1 },
65
- "accent_flag": { "type": "integer", "enum": [0, 1] },
66
- "speed_local_sylps": { "type": ["number", "null"], "minimum": 0 },
67
- "duration_z": { "type": ["number", "null"] }
 
 
 
 
68
  },
69
- "allOf": [
70
- { "if": { "properties": { "end_s": { "type": "number" }, "start_s": { "type": "number" } }, "required": ["start_s","end_s"] },
71
- "then": { "properties": { "end_s": { "minimum": { "$data": "1/start_s" } } } } }
72
- ],
73
- "additionalProperties": false
74
- }
75
- },
76
- "transcript": { "type": "string" }
77
- },
78
- "additionalProperties": false
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
79
  }
 
1
  {
2
  "$schema": "https://json-schema.org/draft/2020-12/schema",
3
+ "$id": "https://aiangel/schema/voice-analysis/1.x",
4
+ "title": "AI Angel Voice Analysis v1.x",
5
+ "description": "Единая схема, принимающая JSON формата 1.0 и 1.1.",
6
+ "anyOf": [
7
+ {
8
+ "title": "v1.0",
 
 
 
 
 
9
  "type": "object",
10
+ "required": ["version", "aggregates", "words", "transcript"],
 
 
 
 
 
 
 
 
 
 
 
 
11
  "properties": {
12
+ "version": { "type": "string", "const": "1.0" },
13
+ "aggregates": {
14
+ "type": "object",
15
+ "required": [
16
+ "duration_sec","voiced_duration_sec","voiced_ratio","rms_dbfs_mean",
17
+ "f0_mean_hz","f0_median_hz","f0_std_hz","f0_stability",
18
+ "words_count","speech_rate_wps","vad"
19
+ ],
20
+ "properties": {
21
+ "duration_sec": { "type": "number", "minimum": 0 },
22
+ "voiced_duration_sec": { "type": "number", "minimum": 0 },
23
+ "voiced_ratio": { "type": "number", "minimum": 0, "maximum": 1 },
24
+ "rms_dbfs_mean": { "type": ["number","null"] },
25
+ "f0_mean_hz": { "type": ["number","null"], "minimum": 0 },
26
+ "f0_median_hz": { "type": ["number","null"], "minimum": 0 },
27
+ "f0_std_hz": { "type": ["number","null"], "minimum": 0 },
28
+ "f0_stability": { "type": ["number","null"], "minimum": 0, "maximum": 1 },
29
+ "words_count": { "type": "integer", "minimum": 0 },
30
+ "speech_rate_wps": { "type": "number", "minimum": 0 },
31
+ "vad": { "type": "string" }
32
+ },
33
+ "additionalProperties": false
34
+ },
35
+ "words": {
36
+ "type": "array",
37
+ "items": {
38
+ "type": "object",
39
+ "required": [
40
+ "start_s","end_s","duration_s","pre_pause_ms","word","prob",
41
+ "f0_mean_hz","f0_z","rms_dbfs","rms_z",
42
+ "accent_score","accent_flag","speed_local_sylps","duration_z"
43
+ ],
44
+ "properties": {
45
+ "start_s": { "type": "number", "minimum": 0 },
46
+ "end_s": { "type": "number", "minimum": 0 },
47
+ "duration_s": { "type": "number", "minimum": 0 },
48
+ "pre_pause_ms": { "type": "integer", "minimum": 0 },
49
+ "word": { "type": "string" },
50
+ "prob": { "type": ["number","null"], "minimum": 0, "maximum": 1 },
51
+ "f0_mean_hz": { "type": ["number","null"], "minimum": 0 },
52
+ "f0_z": { "type": ["number","null"] },
53
+ "rms_dbfs": { "type": ["number","null"] },
54
+ "rms_z": { "type": ["number","null"] },
55
+ "accent_score": { "type": "number", "minimum": 0, "maximum": 1 },
56
+ "accent_flag": { "type": "integer", "enum": [0,1] },
57
+ "speed_local_sylps": { "type": ["number","null"], "minimum": 0 },
58
+ "duration_z": { "type": ["number","null"] }
59
+ },
60
+ "allOf": [
61
+ {
62
+ "if": { "required": ["start_s","end_s"] },
63
+ "then": { "properties": { "end_s": { "minimum": { "$data": "1/start_s" } } } }
64
+ }
65
+ ],
66
+ "additionalProperties": false
67
+ }
68
+ },
69
+ "transcript": { "type": "string" }
70
  },
71
  "additionalProperties": false
72
  },
73
+ {
74
+ "title": "v1.1",
75
+ "type": "object",
76
+ "required": ["version", "aggregates", "words", "transcript"],
77
+ "properties": {
78
+ "version": { "type": "string", "const": "1.1" },
79
+ "aggregates": {
80
+ "type": "object",
81
+ "required": [
82
+ "duration_sec","voiced_duration_sec","voiced_ratio","rms_dbfs_mean",
83
+ "f0_mean_hz","f0_median_hz","f0_std_hz","f0_stability",
84
+ "pitch_confidence_mean",
85
+ "words_count","speech_rate_wps","vad"
86
+ ],
87
+ "properties": {
88
+ "duration_sec": { "type": "number", "minimum": 0 },
89
+ "voiced_duration_sec": { "type": "number", "minimum": 0 },
90
+ "voiced_ratio": { "type": "number", "minimum": 0, "maximum": 1 },
91
+ "rms_dbfs_mean": { "type": ["number","null"] },
92
+ "f0_mean_hz": { "type": ["number","null"], "minimum": 0 },
93
+ "f0_median_hz": { "type": ["number","null"], "minimum": 0 },
94
+ "f0_std_hz": { "type": ["number","null"], "minimum": 0 },
95
+ "f0_stability": { "type": ["number","null"], "minimum": 0, "maximum": 1 },
96
+ "pitch_confidence_mean": { "type": ["number","null"], "minimum": 0, "maximum": 1 },
97
+ "words_count": { "type": "integer", "minimum": 0 },
98
+ "speech_rate_wps": { "type": "number", "minimum": 0 },
99
+ "vad": { "type": "string" }
100
+ },
101
+ "additionalProperties": false
102
  },
103
+ "words": {
104
+ "type": "array",
105
+ "items": {
106
+ "type": "object",
107
+ "required": [
108
+ "start_s","end_s","duration_s","pre_pause_ms","word","prob",
109
+ "f0_mean_hz","f0_z","rms_dbfs","rms_z",
110
+ "accent_score","accent_flag","speed_local_sylps","duration_z",
111
+ "pitch_confidence"
112
+ ],
113
+ "properties": {
114
+ "start_s": { "type": "number", "minimum": 0 },
115
+ "end_s": { "type": "number", "minimum": 0 },
116
+ "duration_s": { "type": "number", "minimum": 0 },
117
+ "pre_pause_ms": { "type": "integer", "minimum": 0 },
118
+ "word": { "type": "string" },
119
+ "prob": { "type": ["number","null"], "minimum": 0, "maximum": 1 },
120
+ "f0_mean_hz": { "type": ["number","null"], "minimum": 0 },
121
+ "f0_z": { "type": ["number","null"] },
122
+ "rms_dbfs": { "type": ["number","null"] },
123
+ "rms_z": { "type": ["number","null"] },
124
+ "accent_score": { "type": "number", "minimum": 0, "maximum": 1 },
125
+ "accent_flag": { "type": "integer", "enum": [0,1] },
126
+ "speed_local_sylps": { "type": ["number","null"], "minimum": 0 },
127
+ "duration_z": { "type": ["number","null"] },
128
+ "pitch_confidence": { "type": ["number","null"], "minimum": 0, "maximum": 1 }
129
+ },
130
+ "allOf": [
131
+ {
132
+ "if": { "required": ["start_s","end_s"] },
133
+ "then": { "properties": { "end_s": { "minimum": { "$data": "1/start_s" } } } }
134
+ }
135
+ ],
136
+ "additionalProperties": false
137
+ }
138
+ },
139
+ "transcript": { "type": "string" }
140
+ },
141
+ "additionalProperties": false
142
+ }
143
+ ]
144
  }