AIAS/3_audio_sdks/ndarray_audio_sdk

NDArray 高级音频工具箱

当前功能包括：

• 获取音频文件的float数组
• 保存音频文件
• 创建给定持续时间和采样率的静音音频段
• 在这个音频样本上加一段静音
• 将任意数量的语音片段连接在一起
• 生成以分贝为单位的音频均方根能量
• 将音频归一化，使其具有所需的有效值(以分贝为单位)
• 用快速傅里叶变换计算线性谱图
• Hanning窗
• 从wav提取mel频谱特征值

运行例子 - AudioExample

...
        NDManager manager = NDManager.newBaseManager(Device.cpu());
        float[] floatArray = AudioArrayUtils.floatData("src/test/resources/test.wav");
        //音频的float数组
        logger.info("音频的float数组: {}", Arrays.toString(floatArray));

        NDArray samples = manager.create(floatArray);
        float rmsDb = AudioUtils.rmsDb(samples);
        //返回以分贝为单位的音频均方根能量
        logger.info("音频均方根能量: {}", rmsDb);

        //提取特征前将音频归一化至-20 dB(以分贝为单位)
        float target_dB = -20f;
        samples = AudioUtils.normalize(samples, target_dB);
        System.out.println("音频归一化: " + samples.toDebugString(1000000000, 1000, 1000, 1000));

        //生成帧的跨步大小(以毫秒为单位)
        float stride_ms = 10f;
        //用于生成帧的窗口大小(毫秒)
        float window_ms = 20f;
        samples = AudioUtils.linearSpecgram(manager, samples, stride_ms, window_ms);
        logger.info("快速傅里叶变换计算线性谱图: {}", samples.toDebugString(1000000000, 1000, 10, 1000));...

命令行查看输出：

...
[INFO ] - 音频均方根能量: -28.989937

[INFO ] - 音频归一化: ND: (134240) cpu() float32
[ 3.09278257e-03,  3.26460390e-03,  1.71821259e-04, ... 133240 more]

[INFO ] - 快速傅里叶变换计算线性谱图: ND: (161, 838) cpu() float32
[[-15.4571, -16.4412, -16.7098, -20.372 , -23.9935, -15.8598, -17.1589, -15.5935, ..., -14.3427],
 [-15.948 , -16.8391, -16.8302, -17.8034, -19.115 , -15.8378, -19.4812, -15.7247, ..., -14.7543],
 [-20.8405, -18.5733, -19.4289, -19.1861, -19.4255, -18.1996, -18.0149, -18.977 , ..., -17.5405],
 [-19.1938, -21.0139, -21.07  , -20.2931, -20.23  , -22.3037, -20.1103, -21.3521, ..., -20.2267],
 [-19.7823, -21.2425, -21.5705, -19.9856, -21.6053, -20.9323, -22.4014, -21.5406, ..., -20.1177],
 [-20.0329, -23.9688, -20.718 , -20.9419, -23.5446, -21.1718, -22.1597, -20.9377, ..., -21.3833],
 [-19.3693, -21.0484, -21.1794, -20.765 , -20.6318, -20.5121, -21.7306, -20.6366, ..., -21.5107],
 [-18.6552, -20.0077, -20.6954, -20.5476, -19.7953, -21.1081, -22.0988, -20.7157, ..., -20.6352],
 [-18.9167, -19.3219, -20.1954, -24.7476, -20.7662, -20.3794, -24.4699, -22.1381, ..., -22.1803],
 [-19.3055, -19.4264, -20.4486, -22.8299, -21.0847, -23.5101, -20.4897, -19.7943, ..., -22.8922],
 ... 151 more]

...

帮助

引擎定制化配置，可以提升首次运行的引擎下载速度，解决外网无法访问或者带宽过低的问题。
引擎定制化配置

官网：

官网链接

Git地址：

Github链接
Gitee链接