Abstract— Speech disorder refers to difficulties in producing vocal sounds or problems related to voice quality. Cases of speech disorders in children and adults have increased manifold, but speech therapists are still hard to be found. An automatic human-machine interface (HMI) is the need of the hour. The proposed system provides a real time feedback to the person with the disorder, which helps them detect mistakes in their speech and rectify them. The HMI incorporates the concept of speech recognition using Linear Predictive Coding technique and provides an audio visual display as feedback to the user. The prototype was simulated using MATLAB and tested with different words.
Keywords— speech therapy, LPC, speech recognition, voice signal
…show more content…
In order to assess the speech intelligibility of dysarthric patients, a system was designed for speech recognition which uses pronunciation confusion network [4].
Speech therapy solutions are also available in an open web platform combined with an online database [5].This initiative uses API’s and the patients do not need to install any software for this.
In contrast to this, a speech processing framework [6] on a reconfigurable chip, has been presented. The chip was a Field Programmable Analog Array (FPAA).The main aim of this framework was noise suppression in the speech signal to enhance speech quality.
Computational problems occur in speech and language processing also. The best way to solve them is by using algorithms that optimise the values under consideration [7]. Features of speech have been matched between the source speaker and the target speaker using deep recurrent neural network [8]. This network consists of recurrent temporal Boltzmann machine and neural network. This framework was
…show more content…
The speech therapy framework was first developed in MATLAB and then incorporated in a minicomputer. In cases of distant talking, environmental factors affect the voice quality. Hence, we need to take measures to ensure the quality of speech signal. Deep neural networks [10] have been used for the acoustic model. It was inferred that with the combination of speech enhancing modules, speech recognition was more accurate. Surveying different concepts, it can be inferred that there is a need to bring up a real time speech therapy framework to help people with speech disorders. This paper deals with a human machine interface which provides real-time feedback to patients through an interactive speech therapy session. OVERVIEW OF THE PROPOSED SYSTEM
Fig.1 shows the block diagram of the proposed system. The audio signal is obtained from the user and it is processed to extract the necessary values.
These values are compared with the signal values of normal speech present in a database. The result of the comparison is given to the user as a graphical display and in the audio form. Fig.1. Block diagram of proposed