kaldi_python_io_read_write_scp_ark_feature_files.py

1import kaldi_python_io
2import numpy as np
3
4# Writing example: Save a dictionary of numpy arrays to a Kaldi archive/script file
5# The 'ark,scp:out.ark,out.scp' format is standard Kaldi syntax
6data_to_write = {
7    "utt_id_1": np.random.randn(100, 40).astype(np.float32),
8    "utt_id_2": np.random.randn(120, 40).astype(np.float32)
9}
10
11with kaldi_python_io.ArchiveWriter('ark,scp:out.ark,out.scp') as writer:
12    for key, matrix in data_to_write.items():
13        writer.write(key, matrix)
14
15# Reading example: Iterate through the created archive using the script file
16with kaldi_python_io.ArchiveReader('out.scp') as reader:
17    for key, matrix in reader:
18        print(f"Read utterance {key} with shape {matrix.shape}")
19
20# Random access example: Get a specific utterance by key
21reader = kaldi_python_io.ArchiveReader('out.scp')
22specific_matrix = reader.get('utt_id_1')
23print(f"Random access successful: {specific_matrix.shape}")
24reader.close()