Rdzleo
93f0e19d1d
ESP32-S3 吊坠设备固件，集成火山引擎 RTC 语音助手、蓝牙配网、 VEML7700 环境光传感器驱动及石头同频匹配交友功能。 VEML7700 驱动： - 基于 ESP-IDF i2c_master API 实现，复用项目 I2cDevice 基类 - 支持 ALS + White 双通道、自动量程、Vishay 非线性校正 - 3 次采样取中位数过滤偶发异常 石头同频匹配算法（双维度）： - 维度1：光谱比值 ALS/White（石头固有光学特征，不随光照强度变化） - 维度2：亮度等级（5级对数划分，排除极端环境差异） - 比值阈值 15%，实测同石头姿势变化波动 1.6%~9.6%，安全余量充足 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
71 lines
3.4 KiB
Python
71 lines
3.4 KiB
Python
# convert audio files to protocol v3 stream
|
||
import librosa
|
||
import opuslib
|
||
import struct
|
||
import sys
|
||
import tqdm
|
||
import numpy as np
|
||
import argparse
|
||
import pyloudnorm as pyln
|
||
|
||
def encode_audio_to_opus(input_file, output_file, target_lufs=None):
|
||
# Load audio file using librosa
|
||
audio, sample_rate = librosa.load(input_file, sr=None, mono=False, dtype=np.float32)
|
||
|
||
# Convert mono to stereo if necessary
|
||
if audio.ndim == 1: # 检查是否为单声道
|
||
audio = np.stack([audio, audio], axis=0) # 复制单声道数据到两个声道，创建立体声
|
||
|
||
if target_lufs is not None:
|
||
print("Note: Automatic loudness adjustment is enabled, which may cause", file=sys.stderr)
|
||
print(" audio distortion. If the input audio has already been ", file=sys.stderr)
|
||
print(" loudness-adjusted or if the input audio is TTS audio, ", file=sys.stderr)
|
||
print(" please use the `-d` parameter to disable loudness adjustment.", file=sys.stderr)
|
||
meter = pyln.Meter(sample_rate)
|
||
current_loudness = meter.integrated_loudness(audio)
|
||
audio = pyln.normalize.loudness(audio, current_loudness, target_lufs)
|
||
print(f"Adjusted loudness: {current_loudness:.1f} LUFS -> {target_lufs} LUFS")
|
||
|
||
# Convert sample rate to 16000Hz if necessary
|
||
target_sample_rate = 16000
|
||
if sample_rate != target_sample_rate:
|
||
audio = librosa.resample(audio, orig_sr=sample_rate, target_sr=target_sample_rate)
|
||
sample_rate = target_sample_rate
|
||
|
||
# Convert audio data back to int16 after processing
|
||
# 对于立体声，需要先调整形状，然后转换为int16
|
||
audio = (audio * 32767).astype(np.int16)
|
||
|
||
# Initialize Opus encoder
|
||
encoder = opuslib.Encoder(sample_rate, 2, opuslib.APPLICATION_AUDIO)
|
||
|
||
# Encode and save
|
||
with open(output_file, 'wb') as f:
|
||
duration = 60 # 60ms per frame
|
||
frame_size = int(sample_rate * duration / 1000)
|
||
# 对于立体声，audio.shape[0]是2（两个声道），audio.shape[1]是采样点数
|
||
for i in tqdm.tqdm(range(0, audio.shape[1] - frame_size, frame_size)):
|
||
# 提取当前帧的两个声道数据
|
||
frame_left = audio[0, i:i + frame_size]
|
||
frame_right = audio[1, i:i + frame_size]
|
||
# 交错两个声道的数据（L1, R1, L2, R2, ...）
|
||
interleaved = np.empty((frame_size * 2,), dtype=np.int16)
|
||
interleaved[0::2] = frame_left
|
||
interleaved[1::2] = frame_right
|
||
# 编码交错的数据
|
||
opus_data = encoder.encode(interleaved.tobytes(), frame_size=frame_size)
|
||
packet = struct.pack('>BBH', 0, 0, len(opus_data)) + opus_data
|
||
f.write(packet)
|
||
|
||
if __name__ == "__main__":
|
||
parser = argparse.ArgumentParser(description='Convert audio to Opus with loudness normalization')
|
||
parser.add_argument('input_file', help='Input audio file')
|
||
parser.add_argument('output_file', help='Output .opus file')
|
||
parser.add_argument('-l', '--lufs', type=float, default=-16.0,
|
||
help='Target loudness in LUFS (default: -16)')
|
||
parser.add_argument('-d', '--disable-loudnorm', action='store_true',
|
||
help='Disable loudness normalization')
|
||
args = parser.parse_args()
|
||
|
||
target_lufs = None if args.disable_loudnorm else args.lufs
|
||
encode_audio_to_opus(args.input_file, args.output_file, target_lufs) |