custom skills - Skills are commands that Alexa can process. They are essentially applications but tailored specifically for AVS. Alexa provides a wide range of custom skills that are readily available to users of Alexa enabled devices. Perhaps the most impressive capability of AVS is that it allows the creation of custom skills that can be used to improve the richness of user experience. The number of existing custom skills runs into the tens of thousands Amazon Alexa Voice Service (AVS): AVS is a platform that allows developers to integrate “Alexa” into their projects. Wi an AVS enabled product, users can stream media, get updates on weather, traffic, and news, as well as ask general knowledge questions. Alexa is triggered when a user says the wake word, “Alexa”. Once the device detects the wake word, it begins to listen for a request. The request is then relayed to cloud based voice service. It is thus a requirement to have a functioning Internet connection. The cloud-based voice service interprets the request and sends a command back to the device or in this case, the smart mirror system. 3.cloud services:-system supports two cloud service Amazon voice service (AVS - Alexa), Amazon web services (AWS), and Google custom search engine. Amazon voice service is used to provide the system’s voice assistant capabilities in addition to the added Alexa skills.Amazon web services is used to create a special voice instructions from the voice assistant. Finally, Google custom search is used in creating a customized search engine to provide multimedia content based on voice requests by the user. The author has created a new skill called “smart mirror” This name represents what the user say to indicate that they want to interact with the skill. The skill is created by visiting the web link https://developer.amazon.com/alexa-skills-kit and selecting to create a new Alexa skill and listing the intents for the new skill. Authors have created a skill name Smart mirror and an intent is created called “ get Time” and a sample phrase like “tell me the time”. If user says “alexa tell mi the time” alexa voice service will invoke the smart mirror skill and it would also recognize users intension for requesting the skill ”’getTime”. There are varinne acnecte while cancidering vnice cammandec Fig.time and event calender TIME & DATE: shows how time and date information is retrieved from the computer system. Fig. to-do-list TO-DO-LIST:it displyas the user shedules and do do list which are costomised by user. Fig .Multimedia content module This is used to display multimedia content like videos, music ,images etc. MULTIMEDIA CONTENCT: Fig.Confidence comparison based on difference frame rates Frame rate refers to no of frame extracted from smart mirror camera and displayed. To find appropriate frame rate and resolution values, the authors enrolled a new user and measured the average confidence achieved for various frame rates and resolution values during authentication. An optimal value, as can be observed across all graphs, would be to select a frame rate of 50 and a resolution of 480 x 360. This is the combination that obtains the lowest average confidence value, 25, while keeping frame processing time low. Fig Average confidence levels of 10 candidates on face recognition candidate | achieved the lowest average confidence of 25. Candidate 2 had the second lowest average confidence of 38. Finally authors decided , a confidence value of 28 to serve as the required threshold for successful facial authentication of the system. Fig. Average confidence levels of 10 candidates on voice recognition Confidence for voice recognition: to find confidence value authors turned off facial recognition and 10 different candidates are said to use the system where 5 candidates are registered and 5 are new. All of the candidates knew the pass phrase and each of them attempted five times. The confidence value of each candidate was calculated based on the average confidence prediction of the voice recognition. Out of 10 five registered candidates (candidates 6 to 10) have much higher confidence value than that of the unregistered users (candidates | to 5). After repeating the experiments among a number of candidates, a confidence value of 83 was selected to serve as the required threshold for successful voice authentication of the system. Based on the threshold values authors have conducted another experiment to verify the authentication accuracy. Authors have created an account on Voicelt API and created different customized phases and users ask to authenticate system.
