|
1 |
package rst.hri;
|
|
2 |
|
|
3 |
import "rst/hri/TextToSpeech.proto";
|
|
4 |
|
|
5 |
option java_outer_classname = "TextToSpeechType";
|
|
6 |
|
|
7 |
/**
|
|
8 |
* Collection of parameters for a Text-to-Speech module.
|
|
9 |
*
|
|
10 |
* @author Soeren Klett <sklett@techfak.uni-bielefeld.de>
|
|
11 |
* @author Birte Carlmeyer <bcarlmey@techfak.uni-bielefeld.de>
|
|
12 |
*/
|
|
13 |
message TextToSpeech {
|
|
14 |
|
|
15 |
/**
|
|
16 |
* Words which should be vocalized.
|
|
17 |
*/
|
|
18 |
optional string text = 1;
|
|
19 |
|
|
20 |
/**
|
|
21 |
* Prosody of spoken words.
|
|
22 |
*/
|
|
23 |
message Prosody {
|
|
24 |
|
|
25 |
/**
|
|
26 |
* The baseline pitch.
|
|
27 |
*/
|
|
28 |
message Pitch {
|
|
29 |
|
|
30 |
/**
|
|
31 |
* Absolute value of pitch.
|
|
32 |
*/
|
|
33 |
optional float absolute = 1;
|
|
34 |
|
|
35 |
/**
|
|
36 |
* Relative change of pitch.
|
|
37 |
*/
|
|
38 |
optional float relative = 2;
|
|
39 |
|
|
40 |
/**
|
|
41 |
* Relative change of pitch in percentage.
|
|
42 |
*/
|
|
43 |
optional float percentage = 3;
|
|
44 |
|
|
45 |
}
|
|
46 |
|
|
47 |
/**
|
|
48 |
* The baseline pitch for the contained words.
|
|
49 |
*/
|
|
50 |
optional Pitch pitch = 1;
|
|
51 |
|
|
52 |
/**
|
|
53 |
* The pitch range (variability).
|
|
54 |
*/
|
|
55 |
message Range {
|
|
56 |
|
|
57 |
/**
|
|
58 |
* Absolute range of pitch in hertz.
|
|
59 |
*/
|
|
60 |
optional float absolute = 1;
|
|
61 |
|
|
62 |
/**
|
|
63 |
* Relative range of pitch.
|
|
64 |
*/
|
|
65 |
optional float relative = 2;
|
|
66 |
|
|
67 |
/**
|
|
68 |
* Relative range of pitch in percentage.
|
|
69 |
*/
|
|
70 |
optional float percentage = 3;
|
|
71 |
|
|
72 |
}
|
|
73 |
|
|
74 |
/**
|
|
75 |
* The pitch range (variability) of the contained words.
|
|
76 |
*/
|
|
77 |
optional Range range = 2;
|
|
78 |
|
|
79 |
/**
|
|
80 |
* The change of volume relative to the baseline.
|
|
81 |
*/
|
|
82 |
message Volume {
|
|
83 |
|
|
84 |
/**
|
|
85 |
* Absolute change of volume.
|
|
86 |
*/
|
|
87 |
optional float absolute = 1;
|
|
88 |
|
|
89 |
/**
|
|
90 |
* Relative change of volume.
|
|
91 |
*/
|
|
92 |
optional float relative = 2;
|
|
93 |
|
|
94 |
/**
|
|
95 |
* Relative change of volume in percentage.
|
|
96 |
*/
|
|
97 |
optional float percentage = 3;
|
|
98 |
|
|
99 |
}
|
|
100 |
|
|
101 |
/**
|
|
102 |
* The desired change of volume of the contained words.
|
|
103 |
*/
|
|
104 |
optional Volume volume = 3;
|
|
105 |
|
|
106 |
/**
|
|
107 |
* A value in milliseconds for the desired time to take to read the contained words.
|
|
108 |
*/
|
|
109 |
optional uint32 duration = 4;
|
|
110 |
|
|
111 |
/**
|
|
112 |
* Relative speech rate.
|
|
113 |
*/
|
|
114 |
optional float rate = 5;
|
|
115 |
|
|
116 |
}
|
|
117 |
|
|
118 |
/**
|
|
119 |
* Prosody to be applied to the given words.
|
|
120 |
*/
|
|
121 |
optional Prosody prosody = 2;
|
|
122 |
|
|
123 |
enum PlaybackOption {
|
|
124 |
|
|
125 |
/**
|
|
126 |
* Play current message.
|
|
127 |
*/
|
|
128 |
PLAY = 0;
|
|
129 |
|
|
130 |
/**
|
|
131 |
* Stop current message.
|
|
132 |
*/
|
|
133 |
STOP = 0;
|
|
134 |
|
|
135 |
/**
|
|
136 |
* PAUSE current message.
|
|
137 |
*/
|
|
138 |
PAUSE = 0;
|
|
139 |
|
|
140 |
/**
|
|
141 |
* RESUME current message.
|
|
142 |
*/
|
|
143 |
RESUME = 0;
|
|
144 |
|
|
145 |
}
|
|
146 |
|
|
147 |
/**
|
|
148 |
* Action to be executed by the Text-to-Speech engine.
|
|
149 |
*/
|
|
150 |
optional PlaybackOption playback_option = 3 [default = PLAY];
|
|
151 |
|
|
152 |
}
|
|
153 |
|
| 0 |
|
-
|