Skip to main content
Log in

On Extending the Noncircular Fast Independent Vector Analysis Algorithm to the Noisy Model

  • Published:
Neural Processing Letters Aims and scope Submit manuscript

Abstract

The noncircular fast independent vector analysis (nc-FastIVA) algorithm is an effective approach to achieve the joint blind source separation (JBSS) problem. However, the original nc-FastIVA algorithm does not take noise into account. In this letter, a new fixed-point algorithm is proposed for more realistic cases when noise is present. Simulations show that the proposed method can have better performance than the original nc-FastIVA algorithm for solving the noisy JBSS problem.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

References

  1. Li Y-O, Adali T, Wang W, Calhoun VD (2009) Joint blind source separation by multiset canonical correlation analysis. IEEE Trans Signal Process 57(10):3918–3929

    Article  MathSciNet  MATH  Google Scholar 

  2. Li X-L, Anderson M, Adali T (2011) Joint blind source separation by generalized joint diagonalization of cumulant matrices. Signal Process 91(10):2314–2322

    Article  MATH  Google Scholar 

  3. Kim T, Attias HT, Lee S-Y, Lee T-W (2007) Blind source separation exploiting higher-order frequency dependencies. IEEE Trans Speech Audio Process 15(1):70–79

    Article  Google Scholar 

  4. Lee I, Kim T, Lee T-W (2007) Fast fixed-point independent vector analysis algorithms for convolutive blind source separation. Signal Process 87(8):1859–1871

    Article  MATH  Google Scholar 

  5. Zhang H-F, Li L-P, Li W-C (2012) Independent vector analysis for convolutive blind noncircular source separation. Signal Process 92(9):2275–2283

    Article  Google Scholar 

  6. Oh M, Park HM (2011) Blind source separation based on independent vector analysis using feed-forward network. Neurocomputing 74(17):3713–3715

    Article  Google Scholar 

  7. Zhang H, Shi Z, Guo C, Feng E (2008) Blind source extraction for noisy mixtures by combining Gaussian moments and generalized autocorrelations. Neural Process Lett 28(3):209–225

    Article  Google Scholar 

  8. Aouada S, Zoubir AM, See CMS (2003) A comparative study on source number detection. In: Signal processing and its applications, proceedings of seventh international symposium, IEEE, vol 1. pp 173–176

  9. Eltoft T, Kim T, Lee T-W (2006) On the multivariate Laplace distribution. IEEE Signal Process Lett 13(5):300–303

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Guobing Qian.

Appendices

Appendix 1: The Proof of (5)

For the right hand side of (4), the first term can be further decomposed as

$$\begin{aligned}&-\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{\hat{\mathbf{x}}}^{[m]}\right| ^{2}} } \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H} {\hat{\mathbf{x}}}^{[k]}} \right] ^{*}{\hat{\mathbf{x}}}^{[k]}} \right\} \nonumber \\&\quad =-\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{x}^{[m]}\right| ^{2}} } \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}} \right] ^{*} \mathbf{x}^{[k]}} \right\} \nonumber \\&\qquad +\,subterm1+subterm2 \end{aligned}$$
(11)

where

$$\begin{aligned}&subterm1 \nonumber \\&\quad =-\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{x}^{[m]}\right| ^{2}} } \right] +\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}\right| ^{2}} \right\} \nonumber \\&\qquad \times \, E\left\{ {\left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right) } \right\} \nonumber \\&\qquad -\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right] +\left| {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right| ^{2}} \right\} \nonumber \\&\qquad \times E\left\{ {\left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}} \right] ^{*}\mathbf{x}^{[k]}} \right\} \end{aligned}$$
(12)
$$\begin{aligned}&subterm2=-E\left\{ \left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right] \right. \nonumber \\&\qquad \left. \times \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}\mathbf{n}^{[k]}} \right) \right\} \end{aligned}$$
(13)

The second term can be further decomposed as

$$\begin{aligned}&\left[ {\mathbf{I}_N +(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}} \right] E\left\{ {\left( \sum _{m=1}^K {|\hat{{y}}_i^{[m]} |^{2}} \right) +|\hat{{y}}_i^{[k]} |^{2}} \right\} \mathbf{w}_i^{[k](n)} \nonumber \\&\quad =E\left\{ {\left( {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{x}^{[m]}\right| ^{2}} } \right) +\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}\right| ^{2}} \right\} \mathbf{w}_i^{[k](n)} \nonumber \\&\qquad +\,subterm3+subterm4 \end{aligned}$$
(14)

where

$$\begin{aligned}&subterm3 \nonumber \\&\quad =E\left\{ {\left( {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}\mathbf{n}^{[m]}\right| ^{2}} } \right) +\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\right| ^{2}} \right\} \mathbf{w}_i^{[k](n)} \nonumber \\&\qquad +\,(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ \left( {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}\mathbf{x}^{[m]}\right| ^{2}} } \right) \right. \nonumber \\&\qquad \left. +\,\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}\right| ^{2} \right\} \mathbf{w}_i^{[k](n)} \end{aligned}$$
(15)
$$\begin{aligned}&subterm4 \nonumber \\&\quad =(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ \left( {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}\mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right) \right. \nonumber \\&\qquad \left. +\,\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\right| ^{2} \right\} \mathbf{w}_i^{[k](n)} \end{aligned}$$
(16)

The third term can be further simplified as

$$\begin{aligned} \begin{array}{l} E\left\{ {{\hat{\mathbf{x}}}^{[k]}\left( {{\hat{\mathbf{x}}}^{[k]}} \right) ^{T}} \right\} \cdot E\left\{ {\left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{\hat{\mathbf{x}}}^{[k]}} \right] ^{2*}} \right\} \mathbf{w}_i^{[k](n)*} \\ \quad =E\left\{ {\mathbf{x}^{[k]}\left( {\mathbf{x}^{[k]}} \right) ^{T}} \right\} E\left\{ {\left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}} \right] ^{2*}} \right\} \mathbf{w}_i^{[k](n)*} \\ \end{array} \end{aligned}$$
(17)

Thus, (4) can be further rewritten as

$$\begin{aligned} \mathbf{w}_i^{[k](n+1)}= & {} -\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{x}^{[m]}\right| ^{2}} } \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}} \right] ^{*}{} \mathbf{x}^{[k]}} \right\} \nonumber \\&+\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{x}^{[m]}\right| ^{2}} } \right] +\left| {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}} \right| ^{2}} \right\} \mathbf{w}_i^{[k](n)} \nonumber \\&+\,E\left\{ {\mathbf{x}^{[k]}\left( {\mathbf{x}^{[k]}} \right) ^{T}} \right\} E\left\{ {\left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{x}^{[k]}} \right] ^{2*}} \right\} \mathbf{w}_i^{[k](n)*} \nonumber \\&+\,f\left( {{\begin{array}{cccc} {\mathbf{w}_i^{[1](n)} }&{} {\mathbf{w}_i^{[2](n)} }&{} \cdots &{} {\mathbf{w}_i^{[K](n)} } \\ \end{array} }} \right) \end{aligned}$$
(18)

where

$$\begin{aligned}&f\left( {{\begin{array}{cccc} {\mathbf{w}_i^{[1](n)} }&{} {\mathbf{w}_i^{[2](n)} }&{} \cdots &{} {\mathbf{w}_i^{[K](n)} } \\ \end{array} }} \right) \\&\quad =-\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}\mathbf{n}^{[k]}} \right) } \right\} \\&\qquad +\,(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ \left( {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right) \right. \nonumber \\&\qquad \left. +\,\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}\mathbf{n}^{[k]}\right| ^{2} \right\} \mathbf{w}_i^{[k](n)} \end{aligned}$$

Appendix 2: The Proof of (8)

Due to the noise of different groups are uncorrelated, it can be obtained that

$$\begin{aligned}&-\,E\left\{ {\left[ {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right) } \right\} \nonumber \\&\quad =-\,E\left\{ {\left[ {\left| {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right| ^{2}} \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right) } \right\} \nonumber \\&\qquad -\,E\left\{ {\left[ {\sum _{m\ne k} {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right] } \right\} \cdot (\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}{} \mathbf{w}_i^{[k](n)} \end{aligned}$$
(19)

and

$$\begin{aligned}&(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ {\left( {\sum _{m=1}^K {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}\mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right) +\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\right| ^{2}} \right\} \mathbf{w}_i^{[k](n)} \nonumber \\&\quad =(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ {2\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}\mathbf{n}^{[k]}\right| ^{2}} \right\} \mathbf{w}_i^{[k](n)} \nonumber \\&\qquad +\,(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ {\left( {\sum _{m\ne k} {\left| \left( {\mathbf{w}_i^{[m](n)} } \right) ^{H}{} \mathbf{V}^{[m]}{} \mathbf{n}^{[m]}\right| ^{2}} } \right) } \right\} \mathbf{w}_i^{[k](n)} \end{aligned}$$
(20)

Thus, (8) can be obtained if

$$\begin{aligned}&-\,E\left\{ {\left[ {\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\right| ^{2}} \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right) } \right\} \nonumber \\&\quad +\,2(\sigma ^{[k]})^{2}{} \mathbf{V}^{[k]}(\mathbf{V}^{[k]})^{H}E\left\{ {\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\right| ^{2}} \right\} \mathbf{w}_i^{[k](n)} =\mathbf{0} \end{aligned}$$
(21)

is true. Next, we will show that the Eq. (21) is true.

Due to the linear combination of independent Gaussian random variables is also Gaussian, each component of \({\bar{\mathbf{n}}}^{[k]}\triangleq \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\) is a Gaussian variable.

Then, \(\hbox {kurt}\left( {\bar{{n}}_i^{[k]} } \right) =E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{4}} \right\} -2\left( {E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{2}} \right\} } \right) ^{2}-E\left\{ {\bar{{n}}_i^{[k]2} } \right\} E\left\{ {\bar{{n}}_i^{[k]*2} } \right\} =0\ (i=1,2,\cdots N)\).

Considering the effect of \(E\left( {{\bar{\mathbf{n}}}^{[k]}{\bar{\mathbf{n}}}^{{[k]}{T}}} \right) =\mathbf{V}^{[k]}E\left( {\mathbf{n}^{[k]}\mathbf{n}^{{[k]}{T}}} \right) \mathbf{V}^{{[k]}{T}}=\mathbf{0}\), we can get \(\hbox {kurt}\left( {\bar{{n}}_i^{[k]} } \right) =E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{4}} \right\} -2\left( {E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{2}} \right\} } \right) ^{2}=0,\) i.e., \(E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{4}} \right\} =2\left( {E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{2}} \right\} } \right) ^{2}\).

Furthermore,

$$\begin{aligned} E\left( {{\bar{\mathbf{n}}}^{[k]}{\bar{\mathbf{n}}}^{{[k]}{H}}} \right)= & {} \mathbf{V}^{[k]}E\left( {\mathbf{n}^{[k]}{} \mathbf{n}^{{[k]}{H}}} \right) \mathbf{V}^{{[k]}{H}} \\= & {} \sigma ^{{[k]}{2}}{\varvec{\Lambda } }^{{[k]}{-1}}\triangleq diag\left( {{\begin{array}{cccc} {E\left\{ {\left| {\bar{{n}}_1^{[k]} } \right| ^{2}} \right\} }&{} {E\left\{ {\left| {\bar{{n}}_2^{[k]} } \right| ^{2}} \right\} }&{} \cdots &{} {E\left\{ {\left| {\bar{{n}}_N^{[k]} } \right| ^{2}} \right\} } \\ \end{array} }} \right) \\ \end{aligned}$$

Thus, for the left hand side of (21), the first term can be rewritten explicitly as

$$\begin{aligned}&-\,E\left\{ {\left[ {\left| \left( {\mathbf{w}_i^{[k](n)} } \right) ^{H} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}\right| ^{2}} \right] \left[ {\left( {\mathbf{w}_i^{[k](n)} } \right) ^{H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right] ^{*}\left( {\mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right) } \right\} \nonumber \\&\quad =-\,E\left\{ {\left| {\mathbf{w}^{[k](n)H} {\bar{\mathbf{n}}}^{[k]}} \right| ^{2}\cdot (\mathbf{w}^{[k](n)H} {\bar{\mathbf{n}}}^{[k]})^{*}\cdot {\bar{\mathbf{n}}}^{[k]}} \right\} \nonumber \\&\quad =-\,E\left\{ {\sum _{a=1}^N {\left( {w_a^{[k](n)*} \bar{{n}}_a^{[k]} } \right) } \cdot \sum _{b=1}^N {\left( {w_b^{[k](n)*} \bar{{n}}_b^{[k]} } \right) ^{*}} \cdot \sum _{c=1}^N {\left( {w_c^{[k](n)*} \bar{{n}}_c^{[k]} } \right) ^{*}} \cdot {\bar{\mathbf{n}}}^{[k]}} \right\} \end{aligned}$$
(22)

The \(i^{th}\) element can be further expressed as

$$\begin{aligned}&-\left| {w_i^{[k](n)} } \right| ^{2}w_i^{[k](n)} E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{4}} \right\} +2w_i^{[k](n)} E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{2}} \right\} E\left\{ {\sum _{l\ne i} {\left| {w_l^{[k](n)} } \right| ^{2}\left| {\bar{{n}}_l^{[k]} } \right| ^{2}} } \right\} \nonumber \\&\quad =-2w_i^{[k](n)} E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{2}} \right\} E\left\{ {\sum _{l=1}^N {\left| {w_l^{[k](n)} } \right| ^{2}\left| {\bar{{n}}_l^{[k]} } \right| ^{2}} } \right\} \end{aligned}$$
(23)

where \(i=1,2\cdots N\).

The second term can be rewritten explicitly as

$$\begin{aligned}&2\sigma ^{[k]2}{} \mathbf{V}^{[k]}{} \mathbf{V}^{[k]H}E\left\{ {\left| {\mathbf{w}^{[k](n)H}{} \mathbf{V}^{[k]}{} \mathbf{n}^{[k]}} \right| ^{2}} \right\} \mathbf{w}^{[k](n)} \nonumber \\&\quad =2\sigma ^{[k]2}{} \mathbf{V}^{[k]}{} \mathbf{V}^{[k]H}E\left\{ {\left| {\mathbf{w}^{[k](n)H}{\bar{\mathbf{n}}}^{[k]}} \right| ^{2}} \right\} \mathbf{w}^{[k](n)} \nonumber \\&\quad =2\sigma ^{[k]2}{} \mathbf{V}^{[k]}{} \mathbf{V}^{[k]H}E\left( {\sum _{l=1}^N {\left| {w_l^{[k](n)} } \right| ^{2}\left| {\bar{{n}}_l^{[k]} } \right| ^{2}} } \right) \mathbf{w}^{[k](n)} \end{aligned}$$
(24)

The ith element can be further expressed as

$$\begin{aligned} 2w_i^{[k](n)} E\left\{ {\left| {\bar{{n}}_i^{[k]} } \right| ^{2}} \right\} E\left\{ {\sum _{l=1}^N {\left| {w_l^{[k](n)} } \right| ^{2}\left| {\bar{{n}}_l^{[k]} } \right| ^{2}} } \right\} \end{aligned}$$
(25)

where \(i=1,2\cdots N\).

Therefore, (21) is true.

Thus, (8) is obtained based on (19), (20) and (21).

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Qian, G., Li, L. On Extending the Noncircular Fast Independent Vector Analysis Algorithm to the Noisy Model. Neural Process Lett 48, 631–642 (2018). https://doi.org/10.1007/s11063-017-9741-0

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11063-017-9741-0

Keywords

Navigation