I’ve met with many device makers who are considering voice but are unsure how to implement in their products. They’re considering far field capability, trigger word naming, and implementation. Having to ponder all of these factors ends up leading to analysis paralysis and months end up passing by with no product on the market to show for the thought effort.
My advice to them is this: start with push. Start with a push to talk feature. It doesn’t require a huge BOM investment (if at all) and can likely use many of the existing components on the device. You don’t need to worry about far field because the user is within arms reach and it doesn’t need a trigger word because the push button is the trigger. If there’s a concern about end of speech detection, this can be relegated by having a push and hold to talk implementation. Push to talk also eliminates certain requirements for implementing Alexa Voice Service, such as having an “Alexa” wake up word. No branding issues.
There is going to be a land rush of voice enabled devices. It’d be better to have a device on the market with reliable voice interaction, even if the experience isn’t hands free like the Echo, than to have nothing and to continue to ponder whether to proceed while others move by.