Enter:~/kaldi/egs/aishell/v1

# address variables (quotation marks can be omitted, but = there must be space between?)
data=/export/a05/xna/data
data_url=www.openslr.org/resources/33

# Load the reference file (use., Similar to the import)
. ./cmd.sh
. ./path.sh

# Set Bash error and exit
set -e # exit on error

# download and decompress the data (should be one of the two)
local/download_and_untar.sh $data $data_url data_aishell
local/download_and_untar.sh $data $data_url resource_aishell
# (DOWNLOAD_AND_UNTAR.SH <data/data download address> <Data output location>)

# Data preparation (should be extracted from WAV data and generate "explanation document"?)
local/aishell_data_prep.sh $data/data_aishell/wav $data/data_aishell/transcript

# extract MFCC features (It is recommended to use a large -capacity disk to store MFCC)
# each WAV is first divided into many frames. Each frame is represented by a series of numbers. MFCC is this string of numbers

mfccdir=mfcc

for x in train test; do

  steps/make_mfcc.sh --cmd "$train_cmd" --nj 10 data/$x exp/make_mfcc/$x $mfccdir
  # MAKE_MFCC.SH <Data Folder> <Log Folder> <MFCC Output Folder>;
  # 12 -Stimemaking after the symbols
  
  sid/compute_vad_decision.sh --nj 10 --cmd "$train_cmd" data/$x exp/make_mfcc/$x $mfccdir
  # compute_vad_decision.sh <<data folder> <Log folder> <MFCC output folder>
  
  utils/fix_data_dir.sh data/$x
  # fix_data_dir.sh requires only one parameter
 
done

# 15 Diag ubm
sid/train_diag_ubm.sh --nj 10 --cmd "$train_cmd" --num-threads 16 \
  data/train 1024 exp/diag_ubm_1024
  
# 1 Full ubm
sid/train_full_ubm.sh --nj 10 --cmd "$train_cmd" data/train \
  exp/diag_ubm_1024 exp/full_ubm_1024
  
# Training Ivector
sid/train_ivector_extractor.sh --cmd "$train_cmd --mem 10G" \
  --num-iters 5 exp/full_ubm_1024/final.ubm data/train \
  exp/extractor_1024

# extract Ivector
sid/extract_ivectors.sh --cmd "$train_cmd" --nj 10 \
  exp/extractor_1024 data/train exp/ivector_train_1024

# Training PLDA
$train_cmd exp/ivector_train_1024/log/plda.log \
  ivector-compute-plda ark:data/train/spk2utt \
  'ark:ivector-normalize-length scp:exp/ivector_train_1024/ivector.scp  ark:- |' \
  exp/ivector_train_1024/plda

# --- ↓ split the test to enroll and eval ↓ ---
mkdir -p data/test/enroll data/test/eval
cp data/test/{
    spk2utt,feats.scp,vad.scp} data/test/enroll
cp data/test/{
    spk2utt,feats.scp,vad.scp} data/test/eval

local/split_data_enroll_eval.py data/test/utt2spk  data/test/enroll/utt2spk  data/test/eval/utt2spk
# split_data_enroll_eval.py 
# 20 20 20 (UTT2SPK) is split into a registration set and verification set
# Input: Test UTT2SPK; Output: Enroll UTT2SPK, Eval UTT2SPK
# For each speaker, the three audio of the lower audio is randomly selected as the verification set, and the remaining other is used as a verification set test

trials=data/test/aishell_speaker_ver.lst
local/produce_trials.py data/test/eval/utt2spk $trials
# local/produce_trials.py
# This script is used to generate trails file: Uttid spkid target | nontarget

utils/fix_data_dir.sh data/test/enroll
utils/fix_data_dir.sh data/test/eval
# utils/fix_data_dir.sh
# This script makes sure that only the segments present in
# all of "feats.scp", "wav.scp" [if present], segments [if present]
# text, and utt2spk are present in any of them.
# It puts the original contents of data-dir into
# data-dir/.backup

# --- ↑ split the test to enroll and eval ↑ ---

# extract the registration set ivector
sid/extract_ivectors.sh --cmd "$train_cmd" --nj 10 \
  exp/extractor_1024 data/test/enroll  exp/ivector_enroll_1024

# extract the Ivector of the verification set
sid/extract_ivectors.sh --cmd "$train_cmd" --nj 10 \
  exp/extractor_1024 data/test/eval  exp/ivector_eval_1024

# Calculate PLDA score

$train_cmd exp/ivector_eval_1024/log/plda_score.log \
  ivector-plda-scoring --num-utts=ark:exp/ivector_enroll_1024/num_utts.ark \
  exp/ivector_train_1024/plda \
  ark:exp/ivector_enroll_1024/spk_ivector.ark \
  "ark:ivector-normalize-length scp:exp/ivector_eval_1024/ivector.scp ark:- |" \
  "cat '$trials' | awk '{print \\\$2, \\\$1}' |" exp/trials_out

# ~kaldi/src/ivectorbin/ivector-plda-scoring.cc
# Use the PLDA model to calculate Trails's matched ratio (OG-Likelihood Ratios)
# Input (Trials-File): "<Key1> <Key2> \ N"
# Output: "<Key1> <Key2> [<Dot-PRODUCT>] \ n"

# Calculate EER
awk '{print $3}' exp/trials_out | paste - $trials | awk '{print $1, $4}' | compute-eer -

# Result
# Scoring against data/test/aishell_speaker_ver.lst
# Equal error rate is 0.140528%, at threshold -12.018

exit 0

1. Linux Shell script some basic knowledge: for example$ variable、<Fu> = <Value>、-Super Porter Settings、. Load
2. kaldi Most of the speaker recognition

Reference:

1. From the speaker recognition DEMO, learn kaldi – (1) run.sh
2. kaldi Introduction: Create the first Chinese ASR (AISHELL-1)