A Neurogram Matching Similarity Index (NMSI) for the Assessment of Audio Quality

Michael Drews, Steffen Schlapak, Stefano Rini, Michele Nicoletti, Werner Hemmert


In this paper, the performance of neurogram and spectrogram similarity indexes for the prediction of perceived audio quality is evaluated. Additionally, a new method for comparison of internal signal representations is introduced that relies on a two-dimensional extension of the Needleman-Wunsch algorithm. It approximates the twodimensional edit distance which is given by the minimum cost to transform one matrix into another by sequentially inserting, deleting and changing the positions and values of single elements. By choosing the cost of each operation, we define a similarity index which can be used for assessment of audio quality among neurograms or spectrograms. We evaluate the performance of this measure to estimate the sound quality of audio files degraded by low bit-rate audio COder-DECoders (CODECs). The new measure shows high correlation with Perceptual Evaluation of Audio Quality (PEAQ) predictions and outperforms other measures of similarity in the literature. We find similarity of spectrograms and neurograms to be sensitive to changes in audio quality for low bit-rate codings.


 DOI: 10.21437/PQS.2013-19

Cite as: Drews, M., Schlapak, S., Rini, S., Nicoletti, M., Hemmert, W. (2013) A Neurogram Matching Similarity Index (NMSI) for the Assessment of Audio Quality. Proc. 4th International Workshop on Perceptual Quality of Systems (PQS 2013), 96-101, DOI: 10.21437/PQS.2013-19.


@inproceedings{Drews2013,
  author={Michael Drews and Steffen Schlapak and Stefano Rini and Michele Nicoletti and Werner Hemmert},
  title={A Neurogram Matching Similarity Index (NMSI) for the Assessment of Audio Quality},
  year=2013,
  booktitle={Proc. 4th International Workshop on Perceptual Quality of Systems (PQS 2013)},
  pages={96--101},
  doi={10.21437/PQS.2013-19},
  url={http://dx.doi.org/10.21437/PQS.2013-19}
}