声纹识别

介绍

声纹识别是一项用计算机程序自动提取说话人特征的技术。

这个 demo 是一个从给定音频文件提取说话人特征，它可以通过使用 PaddleSpeech 的单个命令或 python 中的几行代码来实现。

使用方法

1. 安装

请看安装文档。

你可以从 easy，medium，hard 三中方式中选择一种方式安装。

2. 准备输入

这个 demo 的输入应该是一个 WAV 文件（.wav），并且采样率必须与模型的采样率相同。

可以下载此 demo 的示例音频：

# 该音频的内容是数字串 85236145389
wget -c https://paddlespeech.bj.bcebos.com/vector/audio/85236145389.wav

3. 使用方法

命令行 (推荐使用)

paddlespeech vector --task spk --input 85236145389.wav

echo -e "demo1 85236145389.wav" > vec.job
paddlespeech vector --task spk --input vec.job

echo -e "demo2 85236145389.wav \n demo3 85236145389.wav" | paddlespeech vector --task spk

paddlespeech vector --task score --input "./85236145389.wav ./123456789.wav"

echo -e "demo4 85236145389.wav 85236145389.wav \n demo5 85236145389.wav 123456789.wav" > vec.job
paddlespeech vector --task score --input vec.job

使用方法：

paddlespeech vector --help

参数：

input(必须输入)：用于识别的音频文件。
task (必须输入): 用于指定 vector 处理的具体任务，默认是 spk。
model：声纹任务的模型，默认值：ecapatdnn_voxceleb12。
sample_rate：音频采样率，默认值：16000。
config：声纹任务的参数文件，若不设置则使用预训练模型中的默认配置，默认值：None。
ckpt_path：模型参数文件，若不设置则下载预训练模型使用，默认值：None。
device：执行预测的设备，默认值：当前系统下 paddlepaddle 的默认 device。

输出：

demo  [ -5.749211     9.505463    -8.200284    -5.2075014    5.3940268
    -3.04878      1.611095    10.127234   -10.534177   -15.821609
    1.2032688   -0.35080156   1.2629458  -12.643498    -2.5758228
  -11.343508     2.3385992   -8.719341    14.213509    15.404744
    -0.39327756   6.338786     2.688887     8.7104025   17.469526
    -8.77959      7.0576906    4.648855    -1.3089896  -23.294737
    8.013747    13.891729    -9.926753     5.655307    -5.9422326
  -22.842539     0.6293588  -18.46266    -10.811862     9.8192625
    3.0070958    3.8072643   -2.3861165    3.0821571  -14.739942
    1.7594414   -0.6485091    4.485623     2.0207152    7.264915
    -6.40137     23.63524      2.9711294  -22.708025     9.93719
    20.354511   -10.324688    -0.700492    -8.783211    -5.27593
    15.999649     3.3004563   12.747926    15.429879     4.7849145
    5.6699696   -2.3826702   10.605882     3.9112158    3.1500628
    15.859915    -2.1832209  -23.908653    -6.4799504   -4.5365124
    -9.224193    14.568347   -10.568833     4.982321    -4.342062
    0.0914714   12.645902    -5.74285     -3.2141201   -2.7173362
    -6.680575     0.4757669   -5.035051    -6.7964664   16.865469
  -11.54324      7.681869     0.44475392   9.708182    -8.932846
    0.4123232   -4.361452     1.3948607    9.511665     0.11667654
    2.9079323    6.049952     9.275183   -18.078873     6.2983274
    -0.7500531   -2.725033    -7.6027865    3.3404543    2.990815
    4.010979    11.000591    -2.8873312    7.1352735  -16.79663
    18.495346   -14.293832     7.89578      2.2714825   22.976387
    -4.875734    -3.0836344   -2.9999814   13.751918     6.448228
  -11.924197     2.171869     2.0423572   -6.173772    10.778437
    25.77281     -4.9495463   14.57806      0.3044315    2.6132357
    -7.591999    -2.076944     9.025118     1.7834753   -3.1799617
    -4.9401326   23.465864     5.1685796   -9.018578     9.037825
    -4.4150195    6.859591   -12.274467    -0.88911164   5.186309
    -3.9988663  -13.638606    -9.925445    -0.06329413  -3.6709652
  -12.397416   -12.719869    -1.395601     2.1150916    5.7381287
    -4.4691963   -3.82819     -0.84233856  -1.1604277  -13.490127
    8.731719   -20.778936   -11.495662     5.8033476   -4.752041
    10.833007    -6.717991     4.504732    13.4244375    1.1306485
    7.3435574    1.400918    14.704036    -9.501399     7.2315617
    -6.417456     1.3333273   11.872697    -0.30664724   8.8845
    6.5569253    4.7948146    0.03662816  -8.704245     6.224871
    -3.2701402  -11.508579  ]

Python API

import paddle
from paddlespeech.cli import VectorExecutor

vector_executor = VectorExecutor()
audio_emb = vector_executor(
    model='ecapatdnn_voxceleb12',
    sample_rate=16000,
    config=None,  # Set `config` and `ckpt_path` to None to use pretrained model.
    ckpt_path=None,
    audio_file='./85236145389.wav',
    device=paddle.get_device())
print('Audio embedding Result: \n{}'.format(audio_emb))

test_emb = vector_executor(
    model='ecapatdnn_voxceleb12',
    sample_rate=16000,
    config=None,  # Set `config` and `ckpt_path` to None to use pretrained model.
    ckpt_path=None,
    audio_file='./123456789.wav',
    device=paddle.get_device())
print('Test embedding Result: \n{}'.format(test_emb))
score = vector_executor.get_embeddings_score(audio_emb, test_emb)
print(f"Eembeddings Score: {score}")

输出：

# Vector Result:
 Audio embedding Result:
 [ -5.749211     9.505463    -8.200284    -5.2075014    5.3940268
    -3.04878      1.611095    10.127234   -10.534177   -15.821609
    1.2032688   -0.35080156   1.2629458  -12.643498    -2.5758228
  -11.343508     2.3385992   -8.719341    14.213509    15.404744
    -0.39327756   6.338786     2.688887     8.7104025   17.469526
    -8.77959      7.0576906    4.648855    -1.3089896  -23.294737
    8.013747    13.891729    -9.926753     5.655307    -5.9422326
  -22.842539     0.6293588  -18.46266    -10.811862     9.8192625
    3.0070958    3.8072643   -2.3861165    3.0821571  -14.739942
    1.7594414   -0.6485091    4.485623     2.0207152    7.264915
    -6.40137     23.63524      2.9711294  -22.708025     9.93719
    20.354511   -10.324688    -0.700492    -8.783211    -5.27593
    15.999649     3.3004563   12.747926    15.429879     4.7849145
    5.6699696   -2.3826702   10.605882     3.9112158    3.1500628
    15.859915    -2.1832209  -23.908653    -6.4799504   -4.5365124
    -9.224193    14.568347   -10.568833     4.982321    -4.342062
    0.0914714   12.645902    -5.74285     -3.2141201   -2.7173362
    -6.680575     0.4757669   -5.035051    -6.7964664   16.865469
  -11.54324      7.681869     0.44475392   9.708182    -8.932846
    0.4123232   -4.361452     1.3948607    9.511665     0.11667654
    2.9079323    6.049952     9.275183   -18.078873     6.2983274
    -0.7500531   -2.725033    -7.6027865    3.3404543    2.990815
    4.010979    11.000591    -2.8873312    7.1352735  -16.79663
    18.495346   -14.293832     7.89578      2.2714825   22.976387
    -4.875734    -3.0836344   -2.9999814   13.751918     6.448228
  -11.924197     2.171869     2.0423572   -6.173772    10.778437
    25.77281     -4.9495463   14.57806      0.3044315    2.6132357
    -7.591999    -2.076944     9.025118     1.7834753   -3.1799617
    -4.9401326   23.465864     5.1685796   -9.018578     9.037825
    -4.4150195    6.859591   -12.274467    -0.88911164   5.186309
    -3.9988663  -13.638606    -9.925445    -0.06329413  -3.6709652
  -12.397416   -12.719869    -1.395601     2.1150916    5.7381287
    -4.4691963   -3.82819     -0.84233856  -1.1604277  -13.490127
    8.731719   -20.778936   -11.495662     5.8033476   -4.752041
    10.833007    -6.717991     4.504732    13.4244375    1.1306485
    7.3435574    1.400918    14.704036    -9.501399     7.2315617
    -6.417456     1.3333273   11.872697    -0.30664724   8.8845
    6.5569253    4.7948146    0.03662816  -8.704245     6.224871
    -3.2701402  -11.508579  ]
  # get the test embedding
  Test embedding Result:
  [ -1.9617152    4.2184057   -5.4289927    3.8006616    7.400566
    12.844175     1.4330423    0.4860911  -15.927942   -13.081303
    -4.585545     2.378477     5.5894523  -13.060747    18.578707
    -9.107497    -9.904055     0.7032993    0.7945765   -1.4118854
    -6.4434266   -2.7688267    5.4320455    2.9636188   23.857662
    -4.797293    22.821133    -1.6718386    0.80379957 -10.28131
    -1.0586771    5.840774   -11.794188     0.9715659  -10.794272
    -9.9839325   11.916608   -19.614918    -7.38727     12.361765
  -15.568076     3.796782     1.4648503   -9.617965     1.8912128
    5.5519567    4.1027875    9.565811     1.6652825   -0.06557167
    7.3765106    6.91407     -3.4179301    4.676896     2.4507313
    21.415924    -1.5271066    0.7630236  -15.634208   -24.682417
    12.035311     1.9669697  -13.733474    11.616938   -16.630692
  -16.287516    -7.4265285   -6.4809394    5.4794173   -8.481719
    2.0745668   -7.50969      1.8279544  -15.189501    -4.000386
    -1.5209727    6.975059     4.518711     3.0962887   -6.8465433
    1.3825562    7.6983547   -9.399815    -7.3269534   -2.6540608
    1.3231711    5.0338726   -5.9562182  -10.437971    19.123528
    12.213971    -2.8820174  -20.65914     15.071251     8.114322
    -4.045127     7.5128584   -3.3306584    6.822803    -0.05004288
    -4.4368496   18.926466    14.04377     -5.9657135    4.714744
    10.24277     -3.848245    14.494125     5.3582125   -6.30404
  -14.122616     2.1969411   -5.90989      9.3047      -8.431231
    10.438023   -11.987487    20.954517    -4.279951    -0.3756797
    13.041809    -6.051407   -10.529183     3.7894943   -1.6330183
    6.743382    -0.19549051   7.315633   -19.438568     0.6115422
    4.5697403    2.1208212    0.52282465  -6.9142766   -5.8893275
    0.5135903    0.92921656  -3.0571883   -7.4849505    2.2382743
    -3.0478394    0.08785366   6.810543    -5.1137877   15.182398
    -6.9418297   -8.922732    -2.4528694    7.324874    19.77244
    13.997188    -5.08692    -14.329076    -6.1807523   -1.8777156
    -3.6879017    6.3892293   -3.78877    -13.009837   -16.838747
    -4.1660237   -7.4346085    0.5579437   -2.8482168  -13.509024
    9.329142     8.1292095   -8.064337    -4.002228   -18.78694
    7.7969575  -13.585645    -5.8225474   15.266658    -8.57028
    -7.449079     2.2094946   28.004955    -3.0901644   11.932054
    -1.5897936   -4.826059     6.9232755  -11.169697    -5.235409
    11.251503     2.105524     4.0860977   -0.5384147   19.023642
    1.6203141  -10.608387  ]
  # get the score between enroll and test
  Eembeddings Score: 0.3965281546115875

4.预训练模型

以下是 PaddleSpeech 提供的可以被命令行和 python API 使用的预训练模型列表：

模型	采样率
ecapatdnn_voxceleb12	16k

11 KiB Raw Blame History Unescape Escape