update

main.py

@@ -95,9 +95,12 @@ def click_ring_vad_button(audio: Tuple[int, np.ndarray],
     plt.plot(time, signal / 32768, color="b")
     plt.plot(time, speech_probs, color="gray")
 
+    plt.axhline(y=start_ring_rate, xmin=0.0, xmax=1.0, color="gray", linestyle="-")
+    plt.axhline(y=start_ring_rate, xmin=0.0, xmax=frame_length_ms / 1000 / len(signal) * sample_rate, color="red", linestyle="-")
+
     for start, end in vad_segments:
-        plt.axvline(x=start, ymin=0.15, ymax=0.85, color="g", linestyle="--", label="开始端点")
-        plt.axvline(x=end, ymin=0.15, ymax=0.85, color="r", linestyle="--", label="结束端点")
+        plt.axvline(x=start, ymin=0.15, ymax=0.85, color="g", linestyle="--")
+        plt.axvline(x=end, ymin=0.15, ymax=0.85, color="r", linestyle="--")
 
     temp_image_file = temp_directory / "temp.jpg"
     plt.savefig(temp_image_file)

ring_vad_examples.json

@@ -25,42 +25,42 @@
     ],
     [
         "data/early_media/3300999628164852605.wav",
-        "call_voice", 3, 600, 30, 300, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 300, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/early_media/62/3300999628999191096.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/early_media/62/33009996287818451333.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/noise/00a0a2a3-14ff-4a84-8aee-b18b2fb65355_id-ID_1680237229413.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/noise/000ad44a-fbad-4a22-ba5a-c6dc855779b2_id-ID_1672040947119.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/noise/000da369-6652-4601-b241-33ffbd52a224_id-ID_1676000326981.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/voicemail/00a20d31-e1cb-4c70-821b-6fd151b260ae_id-ID_1671762897272.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/voicemail/000b03b3-172e-4784-8510-24cf37e205ba_id-ID_1672193551438.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/voice/000a3f9a-b2bf-46fd-9c69-477fc62cda51_id-ID_1671935534167.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ],
     [
         "data/call_monitor/id-ID/voice/000cb369-a0ee-44aa-a213-18b036f1baf7_id-ID_1678762306513.wav",
-        "call_voice", 3, 600, 30, 120, 300, 0.4, 0.1, 2.0, 0.3
+        "call_voice", 3, 600, 30, 120, 300, 0.5, 0.2, 2.0, 0.3
     ]
 ]