add sentence service

media_player.py

@@ -38,6 +38,8 @@ from pathlib import Path
 import tkinter as tk
 from tkinter import filedialog, messagebox, ttk
 
+from sentence_analysis import detect_sentence_boundaries
+
 try:
     from PIL import Image as PILImage, ImageTk
 except ImportError:
@@ -71,11 +73,6 @@ LOOP_LABELS = {"off": "循环: 关", "one": "循环: 单曲", "list": "循环:
 SPEED_PRESETS = [0.5, 0.75, 1.0, 1.25, 1.5, 2.0]
 SEEK_STEP = 10
 SEEK_STEP_BIG = 60
-SILENCE_FLOOR_PERCENTILE = 10   # 用能量低分位估计背景噪声底
-SILENCE_THRESHOLD_FACTOR = 1.5  # 静音阈值 = 噪声底 × 系数
-SILENCE_BRIDGE_GAP = 0.06       # 停顿中 60ms 内的短促残响/噪声也按静音处理
-
-
 def fmt_time(ms):
     """把毫秒格式化为 mm:ss 或 h:mm:ss。"""
     if ms is None or ms < 0:
@@ -88,96 +85,6 @@ def fmt_time(ms):
     return f"{minutes:02d}:{seconds:02d}"
 
 
-def detect_sentence_boundaries(path, min_silence=0.30, min_sentence=0.35):
-    """基于音频静音自动切分句子，返回每句话开始的时间（秒）列表。
-
-    语音之间通常有短暂停顿，把停顿前后的连续声音视为一句话。
-    没有音频轨、无法解码或缺少依赖时返回 None。
-    """
-    try:
-        import av
-        import numpy as np
-    except ImportError:
-        return None
-    try:
-        container = av.open(str(path))
-        audio = next((s for s in container.streams if s.type == "audio"), None)
-        if audio is None:
-            container.close()
-            return None
-        sample_rate = 16000
-        resampler = av.AudioResampler(format="fltp", layout="mono", rate=sample_rate)
-        window = 480  # 30ms @ 16kHz
-        frame_seconds = window / sample_rate
-        energies = []
-        buf = []
-
-        def push(data):
-            buf.append(data)
-            total = sum(len(x) for x in buf)
-            if total < window:
-                return
-            arr = np.concatenate(buf)
-            buf.clear()
-            n = len(arr) // window * window
-            win = arr[:n].reshape(-1, window)
-            energies.extend((np.mean(win * win, axis=1) ** 0.5).tolist())
-            if len(arr) > n:
-                buf.append(arr[n:])
-
-        for packet in container.demux(audio):
-            for frame in packet.decode():
-                for out in resampler.resample(frame):
-                    push(out.to_ndarray()[0])
-        for out in resampler.resample(None):
-            push(out.to_ndarray()[0])
-        if buf:
-            push(np.zeros(window, dtype=np.float32))
-        duration = container.duration
-        container.close()
-        if len(energies) < 3:
-            return None
-        energies = np.asarray(energies, dtype=np.float64)
-        floor = float(np.percentile(energies, 95))
-        if floor <= 0.0:
-            return None
-        noise_floor = float(np.percentile(energies, SILENCE_FLOOR_PERCENTILE))
-        threshold = max(0.008, SILENCE_THRESHOLD_FACTOR * noise_floor)
-        silence = energies < threshold
-        bridge_frames = int(round(SILENCE_BRIDGE_GAP / frame_seconds))
-        if bridge_frames > 0:
-            bridged = silence.copy()
-            run_start = None
-            for idx, is_silent in enumerate(silence):
-                if not is_silent and run_start is None:
-                    run_start = idx
-                elif is_silent and run_start is not None:
-                    if run_start > 0 and idx - run_start <= bridge_frames:
-                        bridged[run_start:idx] = True
-                    run_start = None
-            silence = bridged
-        boundaries = [0.0]
-        run_start = None
-        for idx, is_silent in enumerate(silence):
-            if is_silent and run_start is None:
-                run_start = idx
-            elif not is_silent and run_start is not None:
-                if (idx - run_start) * frame_seconds >= min_silence:
-                    end_sec = idx * frame_seconds
-                    if end_sec - boundaries[-1] >= min_sentence:
-                        boundaries.append(end_sec)
-                run_start = None
-        if run_start is not None and (len(silence) - run_start) * frame_seconds >= min_silence:
-            end_sec = len(silence) * frame_seconds
-            if end_sec - boundaries[-1] >= min_sentence:
-                boundaries.append(end_sec)
-        if duration and duration > 0:
-            boundaries = [b for b in boundaries if b < duration / 1e6 - 0.1]
-        return boundaries
-    except Exception:
-        return None
-
-
 class _BaseEngine:
     """播放内核统一接口。"""