K.Imran Shareef N.Kishore Kumar 158P5A0409 (Asst.Professor) IV E.C.E ACEM AGENDA • INTRODUCTION TO AI • WHAT IS SPEECH RECOGNITION? • WHAT DOES IT DO? • TYPES OF SPEECH RECOGNITION • WORKING • STATISTICAL MODELS OF SPEECH RECOGNITION • DISTINCTIVE DECODING METHODS OF SPEECH RECOGNITION • APPLICATIONS • FUTURE SCOPE • CONCLUSION INTRODUCTION TO AI WHAT IS HUMAN INTELLIGENCE? It’s a composition of abilities like: • Learning • Reasoning • Perceiving • Understanding of Language • Feeling WHAT IS ARTIFICIAL INTELLIGENCE? • Basically, “Putting human Intelligence into machines”. • Create intelligent machines and software. • Lots of Innovations. • Systems that act like humans • Systems that think like humans • Systems that think and act rationally TYPES OF AI WEAK AI STRONG AI • Simulates human thoughts and • Matches or exceeds human actions. intelligence. • Actions, decisions and ideas are • Intelligent on their own. programmed into it. • Able to learn freely and adapt, self • All current forms of AI are weak aware. AI. SOFTWARE OF AI • PROLOG(PROgramming in LOgic):All other programming Languages tell the computer how to do something, PROLOG tells the computer what to do. • LISP(LISt Processor): Allows the programmer to arrange the information in orderly sequence. APPLICATIONS OF AI • Speech Recognition. • Facial Recognition. • Military. • Life sciences. • Robotics. • Gaming. WHAT IS SPEECH RECOGNITION? • The process of enabling a computer to identify and respond to the sounds produced in human speech. • Also known as Voice Recognition WHAT DOES IT DO? • Verbal human-machine interaction. • Response from computer in natural language. • User reliability. • Transforms spoken words into text. TYPES OF SPEECH RECOGNITION SPEAKER DEPENDENT SPEAKER INDEPENDENT • Recognizes single person voice. • Recognizes anyone’s voice. • Limited number of words. • Unlimited words. • High accuracy. • Low accuracy. • Specific applications • Numerous applications. WORKING STATISTICAL MODELS OF SPEECH RECOGNITION • ACOUSTIC MODEL • LANGUAGE MODEL • LEXICON MODEL • HIDDEN MARCOV MODEL ACOUSTIC MODEL LANGUAGE MODEL LEXICON MODEL HIDDEN MARKOV MODEL DISTINCTIVE DECODING METHODS OF SPEECH RECOGNITION • Pattern Recognition • Acoustic Phonetic • Artificial Intelligence PATTERN RECOGNITION • Incorporates Pattern comparison and Pattern training. • Utilizes mathematical framework. • Assists in formulating speech patterns. • Further divided into two approaches a) Stochastic approach b) Template approach ACOUSTIC PHONETIC • Assigning labels to sample sounds to recognize sound patterns. • It consists of phonetic units within spoken language. • These units are categorized by collection of acoustic properties. ARTIFICIAL INTELLIGENCE • Combination of Pattern Recognition approach and acoustic phonetic approach. • Utilizes the information related to spectogram,phonetic and linguistic. • Credible and efficient method. • Collects information from respective environment and respond in intelligent manner. SOFTWARES AVAILABLE • Dragon • Ivona • Entrada • Lilyspeech • Braina • Sonix FEW EXAMPLES OF SPEECH RECOGNITION DEVICES ON MOBILE HANDSETS ADVANTAGES • Assists paralyzed and handicapped people. • Saves time for user • Simple handling of software • Comfortable human-machine interaction. • Lower operational costs. DISADVANTAGES • Background noise difficulty. • Different slangs, accent of users. • Voice changeability based on body and environmental condition. • Mixed language. • Difficult to build a perfect system. APPLICATIONS • Telephone speech recognizers for enquiries. • Medical and darkroom appliances. • For handicapped. • Intelligent houses. • Generation of subtitles. • Military and aviation. • In Smart phones. FUTURE SCOPE • Accuracy will become better and better. • Scientists are currently working on a Universal voice recognition translator of sorts, where people of any language can speak, and what they say can be translated into any language, in both speech and text formats. • Though for in the future, it may also be possible for computers to not only recognize what we are saying but understand what we are saying and communicate back with us as well. CONCLUSION • Speech recognition is the process of transforming the input signals (usually speech) into the well-structured sequences of words. • Several techniques and approaches have been developed to overcome this issue. • Amid all of those methods and models, Artificial intelligence is considered as one of the most reliable and adequate approaches. REFERENCES • http://research.microsoft.com/en-us/news/features/speechrecognition- 082911.aspx • http://dl.acm.org/citation.cfm?id=1752355 • http://www.creativecow.net/interstitial.php?url=http%3A%2F%2Fforums. creativecow.net%2Fthread%2F279%2F626&id=0 • www.ijsce.org/attachments/File/v2i5/E1054102512.pdf • http://en.wikipedia.org/wiki/Outline_of_artificial_intelligence • http://www.csd.cs.cmu.edu/research/areas/vis_speech_lang/