!" #$ % 23 Punjabi language is a tonal language belonging to an Indo Aryan language family and has number of speakers all around the world. Punjabi language has gained acceptability in the media & communication and thereby deserves to get a pace in the growing field of automatic speech recognition which has been explored already for number of other Indian and foreign languages successfully. Some work has been done in the field of isolated word and connected word speech recognition for Punjabi language. Acoustic template matching and Vector quantization have been the supporting techniques. Continuous speech recognition is one area where no work has been done so far for Punjabi language. In this paper, an effort has been made to build automatic speech recognizer to recognize continuous speech sentences by using TriPhone based acoustic modeling approach on HTK 3.4.1 speech engine. Overall recognition accuracy has been found to be 82.18% at sentence level and 94.32% at word level. TriPhones, ASR, Hidden Markov Model, MLF, Acoustic Model, HTK, Gaussian Mixtures Automatic speech recognition [1] is a system shown in Fig 1 which performs automatic transformation of an acoustic signal of input speech to a text transcription. The real purpose behind ASR research is to allow a computer to recognize speech in real time with 100% accuracy irrespective of vocabulary size, Recognized word Word Testing Generated Model Training noisy scenario, speaker characteristics, accents and channel conditions. The growing field of automatic speech recognition is facing various issues which are open with due regard to vocabulary size, mode of speech, speaker mode and above all the environmental robustness. These issues are being dealt by researchers so that performance of ASR systems can reach an optimum level. Automatic speech recognition is being explored to great extent in the application areas such as: Voice user interface, Voice interactive response, enhancing social interactive capability of handicapped people, learning a foreign language etc. ! Continuous speech is a speech in natural flow. It is comprised of connected words which are not separated by pauses. Practical implementation of automatic speech recognition requires the capabilities to handle continuous speech. Most impostant feature of continuous speech is that the pronunciations are highly dependent on the context. Recognition of continuous speech is complex due to the following reasons: i. Unlike Isolated word and connected word speech, utterances in continuous speech utterances become overlapped. As a result, word boundaries become unclear and it becomes difficult to identify the start & end points of words. ii. Coarticulation: Coarticulation [2] is a phenomenon which guides natural sounding speech and, brings into effect, the suprasegmental characteristics. It occurs at the boundary between the words. These suprasegmental characteristics degenerate the phonemic boundaries and thereby leads to acoustic variability for the initial & final parts of the word spoken in continuous speech. Finally errors are likely to occur in the recognition system. iii. Rate of speech: Speech rate more specifically signifies the speaking rate and can be determined as number of words spoken per minute. Speech rate always has an impact on our fluency. Style of speaker and nature of text are two important properties of speech rate explained by Mathew [3]. Two parameters based on these properties are as: Mean speech rate, µ = f (Speaker) Variance, σ 2 = g (Cognitive load) Cognitive load is linked with the nature of text and describes the effort made & creativity required for selection of text to be spoken. It has been found that with change in the speech rate [4], duration of vowels show more changes than the consonants. Pre - Processing Spoken Word MFCC Extraction Acoustic Model HMM Model Language Model Pattern Classification Lexical Model