Table 1 Model of S(k, m) and parameters. In Local Gaussian distribution based on NMF, we assume that the variance r(k, m) can be expressed by NMF: \(r(k, m) = \sum _{c} a(c, m) b(c, k)\)

Generalized Gaussian distribution

\(p(S(k, m)) \propto \exp \left( - \left( \frac{|S(k, m)|}{\alpha } \right) ^{\beta } \right)\)

H(k)

Multivariate Laplace distribution

\(p\left( \textbf{S}(m) \right) \propto \exp \left( -\sqrt{\sum _{k}\left| \frac{S(k, m)}{\sigma }\right| ^{2}}\right)\)

H(k)

Local Gaussian distribution based on NMF

\(p\left( S(k, m)\right) \propto \frac{1}{r(k, m)} \exp \left( -\frac{|S(k, m)|^2}{r(k, m)} \right)\)

H(k), a(c, m), b(c, k)