Home Automation by Speech Detection System using Deep Learning

We are now in an age where all new innovations are geared towards helping us make our daily tasks easier. From food to travel, everything can be done by apps. But there is little to no innovation in helping us out at home where it might be needed. The technology exists but it does basic things like switching on a light bulb or a fan using speech recognition. To take it a step further like making routines and making a smart machine to take care of your home and optimize living conditions. Creating a structure for your routine and implementing it in automating your home making sure you optimize your time and money is the goal. To achieve this, macros are used in conjunction with speech recognition with low word error rate. The macros are saved and can be used, changed or removed when required. The macros are used to trigger one or a sequence of computing tasks using one trigger speech. The triggered tasks can range from turning on the lights in a room to multiple things done to wake you up like turning on smooth music, turning on water heater, rolling the blinds up and switching of the fan or ac. These tasks are controllable by the user. New tasks can also be added as new hardware is introduced. An acoustic model comprising of DNN-HMM is used in conjunction with high level network having noise filtration to extract features of speech. The process aims to get better performance compared to existing models.