Building a Rich Arabic Speech Database

2016

Corpus design for speech synthesis is a well-researched topic in languages such as English compared to Modern Standard Arabic, and there is a tendency to focus on methods to automatically generate the orthographic transcript to be recorded (usually greedy methods), which was used in this work. In this work, a study of Modern Standard Arabic (MSA) phonetics and phonology is conducted in order to develop criteria for a greedy method to create a MSA speech corpus transcript for recording. The size of the dataset is reduced a number of times using optimisation methods with different parameters to yield a much smaller dataset with the identical phonetic coverage offered before the reduction. The resulting output transcript is then chosen for recording. A phoneme set and a phonotactic rule-set are created for automatically generating a phonetic transcript of normalised MSA text which is used to annotate and segment the speech corpus after recording, achieving 82.5% boundary precision with some manual alignments (~15% of the corpus) to increase the precision of the automatic alignment. This is part of a larger work to create a completely annotated and segmented speech corpus for MSA speech synthesis with an evaluation of the quality of this speech corpus and, where possible, the quality of each stage in the process.

2014

In this paper, we are interested in the development of an HMM-based Automatic Speech Recognition (ASR) system dedicated to Standard Arabic (SA) language. The acoustical analysis is one of the most delicate tasks in an ASR system development. In order to guarantee enough informative acoustic parameters describing each audio frame, we made a first study on frame windowing (size and period) and a second study on features extraction methods traditionally used in ASR such as MFCC (Melscale Frequency Cepstral Coefficients) and PLP (Perceptual Linear Prediction). This second study allowed us to select the most appropriate coefficients number describing each audio frame. Modeling acoustic models is a key step in any ASR system. That‟s why, we analyzed the effect of varying Gaussian number per HMM state and the number of embedded re-estimations of the Baum-Welch algorithm. To evaluate the proposed ASR system, a multi-speaker SA connected-digits corpus is collected, transcribed and used throu...

This paper describes and proposes an efficient and effective framework for the design and development of a speaker-independent continuous automatic Arabic speech recognition system based on a phonetically rich and balanced speech corpus. The speech corpus contains a total of 415 sentences recorded by 40 (20 male and 20 female) Arabic native speakers from 11 different Arab countries representing the three major regions (Levant, Gulf, and Africa) in the Arab world. The proposed Arabic speech recognition system is based on the Carnegie Mellon University (CMU) Sphinx tools, and the Cambridge HTK tools were also used at some testing stages. The speech engine uses 3-emitting state Hidden Markov Models (HMM) for tri-phone based acoustic models. Based on experimental analysis of about 7 hours of training speech data, the acoustic model is best using continuous observation’s probability model of 16 Gaussian mixture distributions and the state distributions were tied to 500 senones. The language model contains both bi-grams and tri-grams. For similar speakers but different sentences, the system obtained a word recognition accuracy of 92.67% and 93.88% and a Word Error Rate (WER) of 11.27% and 10.07% with and without diacritical marks respectively. For different speakers with similar sentences, the system obtained a word recognition accuracy of 95.92% and 96.29% and a WER of 5.78% and 5.45% with and without diacritical marks respectively. Whereas different speakers and different sentences, the system obtained a word recognition accuracy of 89.08% and 90.23% and a WER of 15.59% and 14.44% with and without diacritical marks respectively.

IEEE access, 2024

Speech recognition is a captivating process that revolutionizes human-computer interactions, allowing us to interact and control machines through spoken commands. The foundation of speech recognition lies in understanding a given language's linguistic and textual characteristics. Although automatic speech recognition (ASR) systems flawlessly convert speech into text for various international languages, their implementation for Arabic remains inadequate. In this research, we diligently explore the current state of Arabic ASR systems and unveil the challenges encountered during their development. We categorize these challenges into two groups: those specific to the Arabic language and those that are more general. Additionally, we propose strategies to overcome these obstacles and emphasize the need for ASR architectures tailored to the Arabic language's unique grammatical and phonetic structure. In addition, we provide a comprehensive and explicit description of various feature extraction methods, language models, and acoustic models utilized in the Arabic ASR system.

Journal of the Franklin …, 2011

This paper presents our work towards developing a new speech corpus for Modern Standard Arabic (MSA), which can be used for implementing and evaluating Arabic speaker-independent, large vocabulary, automatic, and continuous speech recognition systems. The speech corpus was recorded by 40 (20 male and 20 female) Arabic native speakers from 11 countries representing three major regions (Levant, Gulf, and Africa). Three development phases were conducted based on the size of training data, Gaussian mixture distributions, and tied states (senones). Based on our third development phase using 11 hours of training speech data, the acoustic model is composed of 16 Gaussian mixture distributions and the state distributions tied to 300 senones. Using three different data sets, the third development phase obtained 94.32% and 8.10% average word recognition correctness rate and average Word Error Rate (WER), respectively, for same speakers with different sentences (testing sentences). For different speakers with same sentences (training sentences), this work obtained 98.10% and 2.67% average word recognition correctness rate and average WER, respectively, whereas for different speakers with different sentences (testing sentences) this work obtained 93.73% and 8.75% average word recognition correctness rate and average WER, respectively.

2017

Most of the research on speech synthesis or recognize speech has been made on languages like English or French, while many other languages, like Arabic, have not been taken into consideration until recently and sufficient progress has not been made up to now, thus the area of Arabic speech synthesis systems is still in its early development stages. So this work describes a construction of PADAS “Phonetics Arabic Database Automatically segmented” based on a data-driven Markov process. The use of a segmentation database is necessary in speech synthesis and recognizing speech. Manual segmentation is accurate but inconsistent, since it is often produced by more than one label and require time and money. The MAUS segmentation and labeling exist for German speech and other languages but not in Arabic. It is necessary to modify MAUS for establish a segmental database for Arab. The speech corpus contains a total of 600 sentences recorded by 3 (1 female and 2 male) Arabic native speakers fro...

2013

This paper compares Arabic ASR systems for specific dialects versus a system that was trained using pooled data. There are three different dialects in addition to Modern Standard Arabic (MSA). The best Word Error Rate (WER) that could be obtained for the multi-dialect recognition system is 13.7% and the average of the best WERs from the four dialects has been extracted and was found to be 10.2%. Thus, the difference was found to be -3.5%. We compared the use of a multi-dialect system and the use of an accent classifier to extract specific dialect in three different aspects; recognition accuracy, time required and accent classifier accuracy.

Speech recognition for the academic and standard languages has achieved a great development, but in the face of the multiple dialects with their differences, there is still a noticeable lack of recognition rates, especially if these dialects are for non-native speakers of the target language. The challenges facing the recognition or the identification of non-native dialects are numerous, among them the lack of sound databases, whether approved or not.

2020 International Joint Conference on Neural Networks (IJCNN)

This paper presents the design and development of multi-dialect automatic speech recognition for Arabic. Deep neural networks are becoming an effective tool to solve sequential data problems, particularly, adopting an end-to-end training of the system. Arabic speech recognition is a complex task because of the existence of multiple dialects, non-availability of large corpora, and missing vocalization. Thus, the first contribution of this work is the development of a large multi-dialectal corpus with either full or at least partially vocalized transcription. Additionally, the open-source corpus has been gathered from multiple sources that bring non-standard Arabic alphabets in transcription which are normalized by defining a common character-set. The second contribution is the development of a framework to train an acoustic model achieving state-of-the-art performance. The network architecture comprises of a combination of convolutional and recurrent layers. The spectrogram features of the audio data are extracted in the frequency vs time domain and fed in the network. The output frames, produced by the recurrent model, are further trained to align the audio features with its corresponding transcription sequences. The sequence alignment is performed using a beam search decoder with a tetra-gram language model. The proposed system achieved a 14% error rate which outperforms previous systems.

Although considerable research has been devoted to English speech recognition, rather less attention has been paid to Arabic speech recognition. The Arabic language is one of the most commonly used languages worldwide that is in need for accurate audio to text converters. In this paper, we evaluate the recognition performance of the Arabic continuous speech using Soundflower Mac utility. That is, Soundflower was employed as a speaker-independent continuous speech recognition system to evaluate the word error rate (WER) and the accuracy of the Arabic speech. The study also contains a comparative study of the speech recognition performance for male and female native speakers. The experiments conducted using a broadcast news modern standard Arabic (MSA) speech corpus of 2.63 hours (10 male and 10 female speakers). The experimental results show that the accuracy is 54.02 %, and the accuracy of the male and female speakers is almost same. Keywords—Arabic, Speech, Recognition, Corpus, Sou...