I don't know your desired coupling requirements for your system so there are some options. It could be something simple like a 2 lens F# matcher. You should really read this book: "Building Electro-Optical Systems", by Phillip Hobbs. He addresses problems like this in the book.

You should use index-matching fluid/gel

As Sebastian and Peter suggest, a 2-lens coupling scheme is one option. You should aim to match the mode field diameters of the two fibres.

With 5 micron core diameter and numerical aperture 1.7, your tellurite fibre may support multiple modes, but I would expect the mode field diameter of the fundamental mode to be approximately 5 microns. 

MFD for standard single mode fibre is around 10.4 microns, so lens coupling with a magnification between 2x and 2.5x should be suitable.

Pairs of aspheric lenses can be used.  Choose focal lengths in the ratio of the magnification required (shorter focal length next to the MOF), with the least curved glass surface towards the fibre https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=3812

As Boguslaw suggests, there may be a significant Fresnel reflection loss at the tellurite glass / air interface.  Index matching fluid will reduce this, but may also increase spherical aberration with lenses designed for use in air.  There is also a risk that index matching liquid will wick into the interstices of your microstructured fibre, so care is needed - and perhaps some experimentation. 

If interface loss is a problem, an index matching gel may be more appropriate - perhaps in combination with a gradient index lens in place of the aspheric, to eliminate the air gap between fibre and lens.  https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=1209

Ball lenses are another option http://www.edmundoptics.com/resources/application-notes/optics/understanding-ball-lenses/

If your'e looking for a guided-optics solution, you can taper the SMF by heating and stretching it, so it will have the same MFD as your fiber. Then you can splice the two fibers together. You will get some loss from Fresnel reflection (can't add an index matching material this way), but you get a compact connection, without free-space optics.

Ori Golani's suggestion to taper the SMF is good in principle, but is typically used to taper and expand the mode field in  the fibre with smaller MFD.  This is because single mode fibres are generally operated not too far from cut-off, where the MFD of most fibres is close to a minimum.

Reducing the core diameter by stretching the fibre weakens the guidance so that the field spreads further into the cladding.  Conversely, increasing the core diameter, while maintaining the same core-cladding index difference reduces the evanescent field in the cladding, but also takes the fibre into the multimode regime where again the MFD increases with core diameter.

Tapering the outer diameter of the SMF to 5 microns would help in principle, but is probably not a realistic option.