Startups can be made in any area you are passionate in. No matter how big or small the market size is, if you are passionate about an area, feel you can bring something valuable to it, it is absolutely possible to make it a success.
I would start by looking into what areas of mechanical engineering are the most interesting to you, and seeing who customers for those types of products or services are. If you could find a single person willing to pay for what you can do, this can be turned into a startup.
I think the common sense answer would be for you to look at what you're passionate about. I always tell my students that people, in general, tend to be good at what they enjoy, and tend to enjoy what they're good at. However, I think what you're also asking -- or at least should be interested in -- is what the "hot" areas in tech are today. Now, you certainly could invent something of a disruptive innovation and create a new market or service or product. But if you want some hot areas to consider looking for problems to solve in, here they are. First, Internet of Things. It's everywhere. Anything that can be hooked to something that's hooked to the internet or a computer -- which is quite a lot -- is a ripe area for study and for potential business opportunities. Second, Artificial Intelligence. This include Machine Learning and Deep Learning. It also includes open-source tools, and is truly exploding -- far faster than I would have predicted 20 years ago. And be aware of the fact that you don't have to "invent" something in AI. It's like Henry Ford revolutionized the car industry, but he didn't really invent anything in a car. He used technology in new ways. Remember that for every problem you could find, chances are, someone, somewhere, at some time, has already solved it or something very much like it, and you need to find a new way to use that hammer to pound your new kind of nail. And this includes Robotics -- which can make use of AI. Lastly, IOT and AI together make a dynamite combination because what they respectively collect, generate, and use, is Big Data. This is all the stuff that comes from the sensors and channels of IOT and gets processed as a giant image of our world and everything we do in it. Look at those areas, and look for a problem to solve.