HMM speech recognition

Automatic Speech Recognition (ASR) is one of the advanced fields of Natural Language Processing (NLP). Recent past has witnessed valuable research activities in ASR in English, European and East Asian languages. But unfortunately South Asian Languages in general and "Urdu" in particular have received very less attention. In this paper we present an approach to develop an ASR system for Urdu language. The proposed system is based on an open source speech recognition framework called Sphinx4 which uses statistical based approach (HMM: Hidden Markov Model) for developing ASR system. We present a Speaker Independent ASR system for small sized vocabulary, i.e. fifty two isolated most spoken Urdu words and suggest that this research work will form the basis to develop medium and large size vocabulary Urdu speech recognition system.

As speech processing is one of the emerging fields in signal processing, especially because of the great revolution in cellular technologies. It plays great role in the design of interactive systems like speech-to-text converters. So, speech recognizer is a building block in number of systems like speechto-text converters design, text-to-speech converters, and person recognition. We have designed and simulated efficient speech recognizer using linear predictive coding (LPC) and implemented various steps like (training phase, efficient speech models and recognition phase) in Matlab and got all speech recognition coefficients/parameters.

In this paper, we present an approach to develop an automatic speech recognition (ASR) system of Urdu isolated words. Our experimentation is based on a medium vocabulary speech corpus of Urdu, consisting of 250 words. We develop our approach using the open source Sphinx toolkit. Using this platform, we extract the Mel Frequency Cepstral Coefficients (MFCC) features and build a Hidden Markov Model to perform recognition task. We report percentage accuracy for two different experiments based on 100 and 250 words respectively. Experimental results suggest that better recognition accuracy has been achieved with this approach, as compared to the previous results reported on this corpus.

In this paper, we present an approach to develop an automatic speech recognition (ASR) system of Urdu isolated words. Our experimentation is based on a medium vocabulary speech corpus of Urdu, consisting of 250 words. We develop our approach using the open source Sphinx toolkit. Using this platform, we extract the Mel Frequency Cepstral Coefficients (MFCC) features and build a Hidden Markov Model to perform recognition task. We report percentage accuracy for two different experiments based on 100 and 250 words respectively. Experimental results suggest that better recognition accuracy has been achieved with this approach, as compared to the previous results reported on this corpus.

Automatic Speech Recognition (ASR) for Amazigh speech, particularly Moroccan Tarifit accented speech, is a less researched area. This paper focuses on the analysis and evaluation of the first ten Amazigh digits in the noisy conditions from an ASR perspective based on Signal to Noise Ratio (SNR). Our testing experiments were performed under two types of noise and repeated with added environmental noise with various SNR ratios for each kind ranging from 5 to 45 dB. Different formalisms are used to develop a speaker independent Amazigh speech recognition, like Hidden Markov Model (HMMs), Gaussian Mixture Models (GMMs). The experimental results under noisy conditions show that degradation of performance was observed for all digits with different degrees and the rates under car noisy environment are decreased less than grinder conditions with the difference of 2.84% and 8.42% at SNR 5 dB and 25 dB, respectively. Also, we observed that the most affected digits are those which contain the "S" alphabet.

As speech processing is one of the emerging fields in signal processing, especially because of the great revolution in cellular technologies. It plays great role in the design of interactive systems like speech-to-text converters. So, speech recognizer is a building block in number of systems like speechto-text converters design, text-to-speech converters, and person recognition. We have designed and simulated efficient speech recognizer using linear predictive coding (LPC) and implemented various steps like (training phase, efficient speech models and recognition phase) in Matlab and got all speech recognition coefficients/parameters.

Capabilities of domestic service robots could be further improved, if the robot is equipped with an ability to recognize activities performed by humans in its sensory range. For example in a simple scenario a floor cleaning robot can vacuum the kitchen floor after recognizing human activity "cooking in the kitchen". Most of the complex human activities can be sub divided into simple activities which can later used for recognize complex activities. Activities like "take meditation" can be sub divided into simple activities like "opening pill container" and "drinking water". However, even recognizing simple activities are highly challenging due to the similarities between some inter activities and dissimilarities of intra activities which are performed by different people, body poses and orientations. Even a simple human activity like "drinking water" can be performed while the subject is in different body poses like sitting, standing or walking. Therefore building machine learning techniques to recognize human activities with such complexities is non trivial. To address this issue, we propose a human activity recognition technique that uses 3D skeleton features produced by a depth camera. The algorithm incorporates importance weights for skeleton 3D joints according to the activity being performed. This allows the algorithm to ignore the confusing or irrelevant features while relying on informative features. Later these joints were ensembled together to train Dynamic Bayesian Networks (DBN), which is then used to infer human activities based on likelihoods. The proposed activity recognition technique is tested on a publicly available dataset and UTS experiments with overall accuracies of 85% and 90%.

Automatic Speech Recognition (ASR) is one of the advanced fields of Natural Language Processing (NLP). Recent past has witnessed valuable research activities in ASR in English, European and East Asian languages. But unfortunately South Asian Languages in general and "Urdu" in particular have received very less attention. In this paper we present an approach to develop an ASR system for Urdu language. The proposed system is based on an open source speech recognition framework called Sphinx4 which uses statistical based approach (HMM: Hidden Markov Model) for developing ASR system. We present a Speaker Independent ASR system for small sized vocabulary, i.e. fifty two isolated most spoken Urdu words and suggest that this research work will form the basis to develop medium and large size vocabulary Urdu speech recognition system.

Electronic multilingual dictionary is a linguistic resource highly sought after by most natural language processing applications. The construction of such dictionary is a difficult task for under-resourced languages, which generally lack even bilingual dictionaries. This is the case of Moroccan Amazigh language. In this paper, we describe our approach of building automatically a multilingual dictionary for this language via the pivot Universal Networking Language (UNL). The dictionary is based on universal concepts, called universal words in the UNL jargon. The idea consists on merging many bilingual dictionaries for constructing a multilingual dictionary that looks like a matrix, allowing the map between a given word from a language, corresponding to an universal word, to another word from another language corresponding to the same universal word. The advantage of this approach is to build a multilingual dictionary between n languages, even if no bilingual dictionaries exist betwee...

In this paper, we describe the development of a language identification system and a part-of-speech tagger for Latin-Middle English mixed text. To this end, we annotate data with language IDs and Universal POS tags (Petrov et al., 2012). As a classifier, we train a conditional random field classifier for both sub-tasks, including features generated by the TreeTagger models of both languages. The focus lies on both a general and a task-specific evaluation. Moreover, we describe our effort concerning beyond proof-of-concept implementation of tools and towards a more task-oriented approach, showing how to apply our techniques in the context of Humanities research.

Recently increasing attention has been directed to the study of emotional content of speech signals, and hence, many systems have been proposed to identify the emotional content of a spoken utterance. The project of Emotion Recognition from Hindi Speech address to three main aspects of speech recognition system. The first one is the choice of suitable features for speech representation. Using Sparse DTW for feature recognition has improved space efficiency and time complexity. Implementation of automatic emotion recognition system (using MATLAB) provides an accuracy of over 75% for 5 emotions namely: happy, sad, surprise, anger and neutral over a database containing large variety of speakers.

Kopyor coconut is a coconut that has genetic abnormalities which cause the coconut meat to have a unique texture and is detached from the coconut shell. Its uniqueness attracts many enthusiasts resulting in a high economic value, 4-5 times that of the ordinary coconut. From its external appearance, kopyor coconut does not differ with ordinary coconut and this poses a challenge in the detection stage. To date, both farmers and sellers use a traditional approach by listening to the sound of whisk from kopyor coconut to detect them. Unfortunately, this approach relies heavily on experience and expertise of the person. Therefore, a new detection approach is proposed based on sound recognition using Mel Frequency Cepstrum Coefficient (MFCC) as the method for feature extraction and Dynamic Time Warping (DTW) as the method for feature matching. Objects that will be detected are kopyor coconuts and ordinary coconut which has grown mature. By implementing both methods, a program has been dev...

Assistive robots in ambient assisted living environments can be equipped with learning capabilities to effectively learn and execute human activities. This paper proposes a human activity learning (HAL) system for application in assistive robotics. An RGB-depth sensor is used to acquire information of human activities, and a set of statistical, spatial and temporal features for encoding key aspects of human activities are extracted from the acquired information of human activities. Redundant features are removed and the relevant features used in the HAL model. An ensemble of three individual classifiers-support vector machines (SVMs), K-nearest neighbour and random forest-is employed to learn the activities. The performance of the proposed system is improved when compared with the performance of other methods using a single classifier. This approach is evaluated on experimental dataset created for this work and also on a benchmark dataset-the Cornell Activity Dataset (CAD-60). Experimental results show the overall performance achieved by the proposed system is comparable to the state of the art and has the potential to benefit applications in assistive robots for reducing the time spent in learning activities.

Speech recognition is always being an all-time trendy topic for discussion and also for researches and we see a major application in our life. This paper provides the work done on the application of Hidden Markov model to implement isolated word speech recognition on MATLAB and to develop and train the system for set of self-selective words for specific user (user dependent) to get maximum efficiency in word recognition system. Which uses the forward and Baum-welch algorithm and fitting Gaussian of the Baum-welch algorithm for all the iteration perform. We use a sample of 7 alphabets which are recorded in 15 different ways giving total of 105 word to use for training with each word with 15 variations. This system can be used in real world in system security using voice security system and mainly for children and impaired people.

Language is the humans’ most important media of communication and speech is its primary medium. Every person is familiar to his/her mother-tongue from his/her childhood. Thus it gives the ease of communication between human beings. Generally, transfer of information between human and machine is accomplished via keyboard, mouse etc., but human can speak more quickly than typing. Speech input offers high bandwidth information and relative ease of use. Here we deal with the interaction between human and machine via speech recognition. In the current context, Speech Recognition is realized by different approaches. Pattern recognition approach with hidden Markov model is the most used and popular approach by researches as of now. This study was focused on Sinhala speech recognition and objective was to review on the applicability of hidden Markov models as pattern recognition algorithms for Sinhala speech recognition. Applicability is measured by using inclusion and exclusion criteria and search strategy. This will include three main components; Sinhala speech recognition, Hidden Markov Model for speech recognition and Sinhala Language compatibility for HMM. Reviews based on research papers published on above domains have been taken into consideration in search of advances of the Hidden Markov Model approach. Outcomes of the research draw a curtain of various techniques on extended hidden Markov models and show the need of further researches/experiments on hidden Markov model variants for Sinhala speech recognition.

The Linguistic Analysis and the Analysis from the Perspective of Speech Therapy of the Synthetic Method of Elicitation of the Dental Affricative SoundThe author Danuta Pluta-Wojciechowska presents certain reflections on the methods of sound elicitation. She offers the linguistic analysis, as well as, the analysis from the point of view of speech therapy of the synthetic method of sound learning providing, at same time, the example of therapy where professionals dealt with case of dental affricative sound. In the following section of the text, the relevance of linguistic knowledge and the theory of speech therapy in the course of planning the providing speech treatment was stressed.

Praca, której tekst jest tu prezentowana, powstała w 1982 roku i publikowana była w materiałach konferencji, która miała miejsce w 1983 roku. Przedmiotem pracy była dość "egzotyczna" koncepcja rozpoznawania mowy, oparta na tak zwanej rytmiczno-częstotliwościowej analizie sygnału. Była to metoda prosta w realizacji i okazała się przydatna przy rozpoznawaniu ograniczonego słownika wyrazów.

Speech processing is emerged as an essential task in the context of modern communication system. The concept of speech translation deals with the speech signals in a source language A to the target language B. In my work mainly speech to speech translation aiming our local language Telugu to English translation and our national language Hindi to English. In this process, first extraction of features and then reducing noise which can further be used to transfer into text form is done. In this paper, deep learning based modelling technique is employed for speech recognition. After conversion of the text, the data is compared with dictionary data as per the transcriptions for language identification. Mapping is used to generate the signals for transcription. The index value of the recognition is used for language identification. After the language is identified the phonetic approach is used for generate the corresponding text to speech signal.

This paper gives brief information regarding Speech Recognition Technology by including various speech parameters, different approaches to speech recognition, basics of acoustic models, language models, complex algorithms and feature extraction techniques. The objective of paper precisely focusses on speech processing and its applicability in technologies like Human-Machine-Interface (HMI) and various day to day life applications.

Being able to detect and recognize human activities is important for making personal assistant robots useful in performing assistive tasks. The challenge is to develop a system that is low-cost, reliable in unstructured home settings, and also straightforward to use. In this paper, we use a RGBD sensor (Microsoft Kinect) as the input sensor, and present learning algorithms to infer the activities. Our algorithm is based on a hierarchical maximum entropy Markov model (MEMM). It considers a person's activity as composed of a set of sub-activities, and infers the two-layered graph structure using a dynamic programming approach. We test our algorithm on detecting and recognizing twelve different activities performed by four people in different environments, such as a kitchen, a living room, an office , etc., and achieve an average performance of 84.3% when the person was seen before in the training set (and 64.2% when the person was not seen before).

Automatic Speech Recognition (ASR) is one of the advanced fields of Natural Language Processing (NLP). Recent past has witnessed valuable research activities in ASR in English, European and East Asian languages. But unfortunately South Asian Languages in general and " Urdu " in particular have received very less attention. In this paper we present an approach to develop an ASR system for Urdu language. The proposed system is based on an open source speech recognition framework called Sphinx4 which uses statistical based approach (Hidden Markov Model) for developing ASR system. We present a Speaker Independent ASR system for small sized vocabulary, i.e. fifty two isolated most spoken Urdu words and suggest that this research work will form the basis to develop medium and large size vocabulary Urdu speech recognition system.

This paper describes preliminary research results on Polish speech recognition using DTW method aided by HMM. The HMM method is used as a preclassifier searching for n best hypotheses of the recognized words while the DTW – as a precise final classifier. Results of the experiments on large vocabulary isolated word recognition show slight improvement in recognition accuracy compared to alone DTW and HMM methods

This paper describes preliminary research results on speech recognition of speech impaired children. A several Polish phonemes most confusing for speech impaired children were investigated. The records included utterances being the examples of pathological speech. Part of the recorded material was artificially noised by procedure generating white noise. Two most promising types of cepstral coefficients: standard (MFCC) as well as human factor (HFCC) were used for tracking of speech content in frequency domain. For mispronounced phoneme recognition embedded in the word a classical dynamic time warping (DTW) algorithm as well as HMM method were exploited. A phoneme-based approach in DTW method has been proposed. Optimal parameters of HFCC adjusted to the stated recognition task have been found. The superior HFCC performance during conducted recognition experiments especially in strongly noised environment has been observed. The results of the research can be useful for modern logopedic therapy.

W artykule przedstawiono wyniki badań dotyczących automatycznej detekcji wybranych wad wymowy u dzieci za pomocą automatycznego rozpoznawania mowy. Detekcja wady wymowy może być przeprowadzana pod kątem diagnozy lub terapii logopedycznej. Jednym z często spotykanych typów wad wymowy jest substytucja polegająca na podstawianiu prawidłowego fonemu w słowie innym fonemem tego samego języka. W pracy rozważano automatyczną detekcję substytucji pod kątem zastosowań do terapii logopedycznej. W przypadku terapii zadanie automatycznego rozpoznawania mowy upraszcza się do rozpoznawania dwóch głosek: prawidłowej lub nieprawidłowej, ustalonych na podstawie wcześniejszej diagnozy. Jako materiał badawczy przyjęto następujące pary fonemów będące częściami wypowiedzi języka polskiego: {s, sz}, {si, sz}, {c, cz}, {ci, cz}, {dz, drz}, {dzi, drz}. Nagrania pochodziły od dzieci z wadami wymowy oraz od osób, które naśladowały określone wady wymowy. Proces rozpoznawania wady wymowy składał się z dwóch zasadniczych etapów: ekstrakcji cech z sygnału mowy oraz klasyfikacji. W przypadku ekstrakcji cech zbadano dwie metody: standardową metodę MFCC (ang. Mel-frequency cepstral coefficients) oraz stosunkowo niedawno wprowadzoną metodę HFCC (ang. Human-factor cepstral coefficients). Na etapie klasyfikacji przebadano skuteczność rozpoznawania wad wymowy za pomocą czterech metod. Pierwszą badaną metodą była metoda nieliniowej transformacji czasowej (ang. dynamic time warping – DTW). Standardowa metoda DTW jest oparta na modelach całych słów. W przypadku rozważanego problemu terapii substytucji słowa są rozróżnianie jedynie za pomocą jednego fonemu. W takiej sytuacji standardowa metoda DTW często zawodzi zwłaszcza, że fonemy rozróżniające dwa słowa są najczęściej podobne do siebie akustycznie. Ponadto segmenty poza obszarem rozróżniających słowa fonemów często podlegają różnym zniekształceniom lub zakłóceniom, co może dać w rezultacie większą niż zwykle odległość DTW pomiędzy słowami tej samej klasy. Zaproponowano modyfikację standardowej metody DTW polegającą na obliczaniu odległości DTW jedynie między fonemami będącymi składowymi danego słowa. W proponowanym rozwiązaniu zakłada się, że klasa rozpoznawanego słowa jest znana oraz, że słowo może być wypowiedziane prawidłowo lub nieprawidłowo zgodnie z postawioną wcześniej diagnozą. Oprócz metod opartych na nieliniowej transformacji czasowej zbadano również metodę klasyfikacji bazującą na niejawnych modelach Markowa (ang. hidden Markov models – HMM). Metodę HMM badano dla modeli całych słów oraz dla modeli fonemów. W przeprowadzonych badaniach zaobserwowano następujące tendencje: • metoda DTW rozpoznająca fonemy dała wyższe skuteczności niż metoda DTW rozpoznająca słowa • metoda HMM dla modeli fonemów dała lepsze rezultaty w porównaniu z metodą HMM dla całych słów • Skuteczności rozpoznawania w oparciu o cechy HFCC były wyższe w porównaniu ze standardowymi cechami MFCC • W porównaniu z klasyfikatorem DTW, metoda HMM dała nieznacznie gorsze wyniki, jednak problem ten wymaga dalszych badań. Opracowane metody rozpoznawania mogą znaleźć zastosowanie w diagnostyce i terapii wad wymowy u dzieci. W szczególności można z ich wykorzystaniem dokonywać wykrywania substytucji następujących par fonemów: sz-s, cz-c oraz drz-dz do celów terapii logopedycznej. Dla par fonemów sz-si, cz-ci, drz-dzi muszą zostać opracowane inne, bardziej skuteczne metody. Potencjalne przyszłe kierunki badawcze obejmują zastosowanie metody analizy składowych głównych oraz analizy dyskryminacyjnej zarówno w metodzie DTW jak i HMM. Planuje się również dalszą optymalizację parametrów metody HMM. Opisywane w pracy badania były sponsorowane z grantu MNiI nr 1 H01F 046 28.

Praca jest przedstawiana w postaci maszynopisu, ale była ona publikowana - najpierw w materiałach konferencyjnych, a potem w czasopiśmie. Opisane są w niej właściwości sygnału mowy, techniki jego przetwarzania przy użyciu komputerów, sposoby parametryzacji oraz metody rozpoznawania.

Automotive theft has been a concern in Nigeria and greater challenge stems from the inability of the security agencies to retrieve stolen vehicles through forensic mechanisms. The need to protect automobiles from being stolen without any leads is very important. Prevention of vehicles from unsolved theft can be done by developing forensic aided systems in such vehicles. This paper aims at proposing an automobile forensic aided system, through the application of Face Recognition to identify an unauthorized driver. A comparison is made through the database of enrolled faces for verification. The system captures the face of the driver and checks for verification, if no match is found it detects the identity as an imposter then forwards the facial biometric data of the driver to a remote storage for further investigation. The Principal Component Analysis (PCA) algorithm is deployed to design for facial recognition and identification.

Human activity recognition is crucial for intuitive cooperation between humans and robots. We present an approach for activity recognition for applications in the context of human-robot interaction in industrial settings. The approach is based on spatial and temporal features derived from skeletal data of human workers performing assembly tasks. These features were used to train a machine learning framework, which classifies discrete time frames with Random Forests and subsequently models temporal dependencies between the resulting states with a Hidden Markov Model. We considered the following three groups of activities: Movement, Gestures, and Object handling. A dataset has been collected which is comprised of 24 recordings of several human workers performing such activities in a human-robot interaction environment, as typically seen at small and medium-sized enterprises. The evaluation shows that the approach achieves a recognition accuracy of up to 88% for some activities and an average accuracy of 73%.

In this work, we propose an approach that relies on cues from depth perception from RGB-D images, where features related to human body motion (3D skeleton features) are used on multiple learning classifiers in order to recognize human activities on a benchmark dataset. A Dynamic Bayesian Mixture Model (DBMM) is designed to combine multiple classifier likelihoods into a single form, assigning weights (by an uncertainty measure) to counterbalance the likelihoods as a posterior probability. Temporal information is incorporated in the DBMM by means of prior probabilities, taking into consideration previous probabilistic inference to reinforce current-frame classification. The publicly available Cornell Activity Dataset [1] with 12 different human activities was used to evaluate the proposed approach. Reported results on testing dataset show that our approach overcomes state of the art methods in terms of precision, recall and overall accuracy. The developed work allows the use of activities classification for applications where the human behaviour recognition is important, such as human-robot interaction, assisted living for elderly care, among others.

In this work, we propose an approach that relies on cues from depth perception from RGB-D images, where features related to human body motion (3D skeleton features) are used on multiple learning classifiers in order to recognize human activities on a benchmark dataset. A Dynamic Bayesian Mixture Model (DBMM) is designed to combine multiple classifier likelihoods into a single form, assigning weights (by an uncertainty measure) to counterbalance the likelihoods as a posterior probability. Temporal information is incorporated in the DBMM by means of prior probabilities, taking into consideration previous probabilistic inference to reinforce current-frame classification. The publicly available Cornell Activity Dataset [1] with 12 different human activities was used to evaluate the proposed approach. Reported results on testing dataset show that our approach overcomes state of the art methods in terms of precision, recall and overall accuracy. The developed work allows the use of activities classification for applications where the human behaviour recognition is important, such as human-robot interaction, assisted living for elderly care, among others.

The  increase  of  video  lecture  data  on  World  Wide  Web  is  rapid  therefore  an  efficient  method  of  data  retrieval  is
needed.  So  the  system  providing  a  method  for  data  retrieval  from  the  lecture  video  is  impleme
nted  which  will
extract  the  speech
d
ata.  Automatic  video  segmentation  is  applied  first.
Automatic  Speech  Recognition  (ASR)
applied
on the lecture video tracks that are the speech contents given by lecturer during the video.
The extracted text
data is  saved  in the  form of templates  for future reference.
The 80
.97
% of accuracy  in recognising the  letters from
lecture videos is achieved and the extracted information is saved so that the quality of learning is improved

Language is the most important means of communication
and speech is its main medium. In human to machine
interface, speech signal is transformed into analog and
digital wave form which can be understood by machine.
Speech technologies are vastly used and has unlimited
uses. These technologies enable machines to respond
correctly and reliably to human voices, and provide useful
and valuable services .This paper gives an overview of the
speech recognition process, its basic model, and its
application, approaches and also discuss comparative
study of different approaches which are used for speech
recognition system. These papers also give an overview
of different techniques of speech recognition system to
summarize some of the well known methods used in
various stages of speech recognition system.

Capabilities of domestic service robots could be further improved, if the
robot is equipped with an ability to recognize activities performed by humans in its
sensory range. For example in a simple scenario a floor cleaning robot can vacuum
the kitchen floor after recognizing human activity “cooking in the kitchen”. Most
of the complex human activities can be sub divided into simple activities which
can later used for recognize complex activities. Activities like “take meditation”
can be sub divided into simple activities like “opening pill container” and “drinking
water”. However, even recognizing simple activities are highly challenging due to
the similarities between some inter activities and dissimilarities of intra activities
which are performed by different people, body poses and orientations. Even a sim-
ple human activity like “drinking water” can be performed while the subject is in
different body poses like sitting, standing or walking. Therefore building machine
learning techniques to recognize human activities with such complexities is non
trivial. To address this issue, we propose a human activity recognition technique
that uses 3D skeleton features produced by a depth camera. The algorithm incor-
porates importance weights for skeleton 3D joints according to the activity being
performed. This allows the algorithm to ignore the confusing or irrelevant features
while relying on informative features. Later these joints were ensembled together to
train Dynamic Bayesian Networks (DBN), which is then used to infer human activi-
ties based on likelihoods. The proposed activity recognition technique is tested on a
publicly available dataset and UTS experiments with overall accuracies of 85% and
90%

Ability to recognize human activities will enhance the capabilities of a robot that interacts with humans. However automatic detection of human activities could be challenging due to the individual nature of the activities. In this paper, we present human activity detection model that uses only 3-D skeleton features generated from an RGB-D sensor (Microsoft KinectTM). To infer the human activities, we implemented Gaussian Mixture Modal (GMM) based Hidden Markov Model(HMM). GM outputs of the HMM were effectively able to capture multimodel nature of 3D positions of each skeleton joint. We test our model in a publicly available data-set that consists of twelve different daily activities performed by four different people. The proposed model recorded recognition recall accuracy of 84% with previously seen people and 78% with previously unseen people.

The last ten years have been the witnesses of the emergence of any kind of video content. Moreover, the appearance of dedicated websites for this phenomenon has increased the importance the public gives to it. In the same time, certain individuals are deaf and occasionally cannot understand the meanings of such videos because there is not any text transcription available. Therefore, it is necessary to find solutions for the purpose of making these media artefacts accessible for most people. Several software propose utilities to create subtitles for videos but all require an extensive participation of the user. Thence, a more automated concept is envisaged. This thesis report indicates a way to generate subtitles following standards by using speech recognition. Three parts are distinguished. The first one consists in separating audio from video and converting the audio in suitable format if necessary. The second phase proceeds to the recognition of speech contained in the audio. The ultimate stage generates a subtitle file from the recognition results of the previous step. Directions of implementation have been proposed for the three distinct modules. The experiment results have not done enough satisfaction and adjustments have to be realized for further work. Decoding parallelization, use of well trained models, and punctuation insertion are some of the improvements to be done.