Nearest Neighbor Search-Based Bitwise Source Separation Using Discriminant Winner-Take-All Hashing

We propose an iteration-free source separation algorithm based on Winner-Take-All (WTA) hash codes, which is a faster, yet accurate alternative to a complex machine learning model for single-channel source separation in a resource-constrained environment. We first generate random permutations with WTA hashing to encode the shape of the multidimensional audio spectrum to a reduced bitstring representation. A nearest neighbor search on the hash codes of an incoming noisy spectrum as the query string results in the closest matches among the hashed mixture spectra. Using the indices of the matching frames, we obtain the corresponding ideal binary mask vectors for denoising. Since both the training data and the search operation are bitwise, the procedure can be done efficiently in hardware implementations. Experimental results show that the WTA hash codes are discriminant and provide an affordable dictionary search mechanism that leads to a competent performance compared to a comprehensive model and oracle masking.

• Publication year

  2019

• Venue

  arXiv.org

• Publication date

  2019-08-26

• Fields of study

  Computer Science, Engineering

• Identifiers
  arXiv 1908.09799
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Bitwise Source Separation on Hashed Spectra: An Efficient Posterior Estimation Scheme Using Partial Rank Order Metrics

  2017cited by this paper
• A Wavenet for Speech Denoising

  2017cited by this paper
• Single channel audio source separation using convolutional denoising autoencoders

  2017cited by this paper
• Monaural Singing Voice Separation with Skip-Filtering Connections and Recurrent Inference of Time-Frequency Mask

  2017cited by this paper
• Deep clustering: Discriminative embeddings for segmentation and separation

  2015influential reference
• Efficient manifold preserving audio source separation using locality sensitive hashing

  2015cited by this paper
• An Experimental Study on Speech Enhancement Based on Deep Neural Networks

  2014cited by this paper
• GloVe: Global Vectors for Word Representation

  2014cited by this paper
• Fast, Accurate Detection of 100,000 Object Classes on a Single Machine

  2013cited by this paper
• Manifold Preserving Hierarchical Topic Models for Quantization and Approximation

  2013cited by this paper
• Towards Scaling Up Classification-Based Speech Separation

  2013cited by this paper
• Distributed Representations of Words and Phrases and their Compositionality

  2013cited by this paper
• Deep neural networks for single channel source separation

  2013cited by this paper
• Universal speech models for speaker independent single channel source separation

  2013cited by this paper
• Online PLCA for Real-Time Semi-supervised Source Separation

  2012cited by this paper
• Vocal separation using nearest neighbours and median filtering

  2012cited by this paper
• The power of comparative reasoning

  2011cited by this paper
• A short-time objective intelligibility measure for time-frequency weighted noisy speech

  2010influential reference
• A Sparse Non-Parametric Approach for Single Channel Separation of Known Sounds

  2009cited by this paper
• Speech denoising using nonnegative matrix factorization with priors

  2008cited by this paper
• Monaural Sound Source Separation by Nonnegative Matrix Factorization With Temporal Continuity and Sparseness Criteria

  2007cited by this paper
• Learning Spectral Clustering, With Application To Speech Separation

  2006cited by this paper
• Performance measurement in blind audio source separation

  2006influential reference
• On the approximate W-disjoint orthogonality of speech

  2002cited by this paper
• The auditory organization of speech and other sources in listeners and computational models

  2001cited by this paper
• Similarity Search in High Dimensions via Hashing

  1999cited by this paper
• Signature Verification Using A "Siamese" Time Delay Neural Network

  1993cited by this paper
• 1989 IEEE ASSP Workshop on Applications of Signal Processing to Audio and Acoustics

  1990cited by this paper

• No citing papers are available for this paper.