Role of neural network models for developing speech systems

Abstract

This paper discusses the application of neural networks for developing different speech systems. Prosodic parameters of speech at syllable level depend on positional, contextual and phonological features of the syllables. In this paper, neural networks are explored to model the prosodic parameters of the syllables from their positional, contextual and phonological features. The prosodic parameters considered in this work are duration and sequence of pitch (F 0 ) values of the syllables. These prosody models are further examined for applications such as text to speech synthesis, speech recognition, speaker recognition and language identification. Neural network models in voice conversion system are explored for capturing the mapping functions between source and target speakers at source, system and prosodic levels. We have also used neural network models for characterizing the emotions present in speech. For identification of dialects in Hindi, neural network models are used to capture the dialect specific information from spectral and prosodic features of speech.

Processamento auditivo central: implicações para o processo tradutório do Português para a Língua Brasileira de Sinais

“…This result indicated that G2 had an inability to identify the duration of sounds, the non-verbal sounds in the presence of other sounds, and the acquisition of information about the tone of the language (8,9) , Such temporal aspects of speech are fundamental to understanding the information, and act as a resource linked to the processing of language subsystems, phonological aspects, and syntactic and semantic cues. Research in the field of psycholinguistics provide enough data for such findings (23,24,25,26) . The literature indicates that emphasis should be placed on paralinguistic aspects (non-verbal) and on the language pragmatics, as they facilitate the enunciation of the speaker, and allow for different interpretations of the message's meanings.…”

“…It can be divided into: its linguistic aspects, which include speech, as well as its extralinguistic aspects, or paralinguistic elements (stress/loudness, pitch and speech intonation); non-verbal communication (gestures, facial and body language and eye contact); and metalinguistic aspects (ability to use language to analyze language) (23) . The combination of paralinguistic aspects constitutes the term oral/auditory speech prosody, which is responsible for conveying the emotional and inferential perceptions of speech that contribute largely towards speech intelligibility for the listener (24,25) . The interpreters of G1 and G2 went through a process of understanding the spoken language of the proposed audio, for five minutes.…”

“…The interpreters of G1 and G2 went through a process of understanding the spoken language of the proposed audio, for five minutes. Both groups were exposed to the interpretation of the linguistic and paralinguistic content of the listened speech signal, which were aspects that were not excluded in this discussion despite its absence of eye contact or complementary gestures in the listened speech of a speaker, because them were also aspects of language, these aspects were also present (25) .…”

Acoustic-Phonetic Feature Based Dialect Identification in Hindi Speech

“…These features have been proven to be the key feature in human perception of speech. In [9] it is shown that combination of spectral and prosodic features can improve the system performance. Few prosodic features have been investigated here in combination with spectral features for their capability to give dialect specific information.…”

Filterbank optimization for robust ASR using GA and PSO

“…The second layer (first hidden layer) of the network has more units than the input layer, and it can be interpreted as capturing some local features in the input space. The third layer (second hidden layer) has fewer units than the first layer, and can be interpreted as capturing some global features (Rao 2011). Thus the activation function for the units at the input layer is linear, and for the units at the hidden layers, it is non-linear.…”