Novel denoising approach to neuronal spike signals recorded by microelectrode array

Dan Wu, Zhou Yan Feng, Jing Wang

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Based on the fact that there are strong spatial correlations among the noises recorded in multi-channels in a microelectrode array, a new denoising method was developed by combining principle component analysis (PCA) with wavelet threshold method, in order to eliminate various types of noises in extracellular single neuronal action potential (i.e. spike) recordings. The largest noise component in the multi-channel recording signals was first extracted by using PCA decomposition and was removed from the raw signals. The signals then went through wavelet multi-level decomposition. The residual noises in every wavelet level were removed by a soft-thresholding method. Both simulation data and experimental recordings were used to test this PCA-wavelet combined algorithm. The results showed that the algorithm can simultaneously suppress white noise and colored noise, and significantly increase the signal-to-noise ratio of spike signals. Especially for the multiple channel recordings with independent spike signals and highly correlated noises, the performance of the PCA-wavelet combined algorithm significantly surmounts the individual performance of PCA denoising and wavelet threshold denoising used separately. Therefore, the novel PCA-wavelet combined algorithm provides an effective and useful method to denoise multichannel spike signals.

Original languageEnglish (US)
Pages (from-to)104-110
Number of pages7
JournalZhejiang Daxue Xuebao(Gongxue Ban)/Journal of Zhejiang University (Engineering Science)
Issue number1
StatePublished - Jan 2010
Externally publishedYes


  • Microelectrode array
  • Principal component analysis
  • Signal-to-noise ratio
  • Spike
  • Wavelet threshold denoising


Dive into the research topics of 'Novel denoising approach to neuronal spike signals recorded by microelectrode array'. Together they form a unique fingerprint.

Cite this