Evaluation Of Swallowing And Phonation Functions Based On Neuromuscular Information

Swallowing is the most basic physiological function to maintain human survival and to provide essential nutrients for human life.Meanwhile,phonation is the most important physiological function to maintain social communication and plays an important role to communicate with the outside world.The normal maintenance of swallowing and phonation functions depends on the movement and coordination of facial and neck muscles.Once these muscles are damaged,the swallowing and phonation functions will be affected by varying degrees,causing swallowing disorders or dysphonia.The surface electromyography(sEMG)signal is a weak voltage signal collected by electrodes placed on the surface of the skin.sEMG signal has a strong correlation with the activity and functional state of the muscle,and can reflect the activity level of the corresponding neuromuscular activities.Currently,most of the clinical and research settings that are related to swallowing and pronunciation often use a small number of electrodes to collect sEMG signals from facial or neck muscles.Due to the limited amount of information with the lmited electrodes,it is difficult to make a comprehensive understanding and evaluation of swallowing and pronunciation functions,resulting in inability to accurately capture the damaged muscle groups that cause abnormal swallowing or pronunciation.Therefore,effective functional rehabilitation training for the damaged muscle groups cannot be realized,thereby limiting the clinical rehabilitation outcomes of swallowing and pronunciation disorders.In view of the problems existing in the current theoretical research and clinical applications of swallowing and pronunciation functions,this study proposed to use a high-density surface electromyography technology to evaluate the electrophysiological functions and cooperative working mechanisms of the facial and neck muscles during swallowing and pronunciation.This study is useful to develop an evaluation system to provide more accurate,non-invasive and objective clues for the clinical diagnosis of patients with dysphagia and dysarthria.In the swallowing experiment,96-channel surface EMG electrodes were used,and the electrodes were distributed in the form of a 6×16 two-dimensional array on the subject's neck to obtain physiological information.The subjects completed different tasks when swallowing objects of different size and viscosity with different head postures.The high-density sEMG signal of the entire swallowing process was synchronously recorded using the Dutch TMSi-128(Refa-136ch)high-performance electromyographic acquisition system.The experimental results showed that the highdensity sEMG energy map was closely related to the physiological and biomechanical process of swallowing.It could also help to dynamically visualize the entire swallowing process and help to demonstrate spatial distribution of sEMG energies during the swallowing process.It was also observed that the dry swallowing process seems to take longer time than liquid swallowing.As the viscosity of the swallowed object increased,the maximum activity of the sEMG signal in the electrode coverage area was significantly enlarged.When the head of the subject was kept in the middle position,the muscle activities on the left and right sides were almost symmetrical among different swallowing tasks.However,if the head was tilted to one side,the sEMG energy map showed that the major energy of muscle activity would shift to the corresponding side.The high-density sEMG method proposed in this study provides a new method for analyzing muscle contraction and coordination during swallowing,and may provide a practical tool for evaluating the normality of muscle function.It can also provide experimental and theoretical basis for locating the possible causes of patients with dysphagia.In the phonation experiment,120 surface electrodes were arranged in an array to cover all the facial and neck muscles related to phonation.80 of these electrodes were evenly distributed on the neck in a 5×16 array,and the other 40 electrodes were divided into two groups,with two 4×5 arrays symmetrically placed on the left and right sides of the facial and neck muscles.In a relaxed state,the subjects were asked to complete the phonation tasks of 3 vowels([a:],[i:],[?:])with different sound intensities and pitches.They were also asked to speak the numbers from 0 to 9 in both Chinese and English respectively,and the acoustic speech signals and high-density sEMG data were synchronously collected in all phonation tasks.For the digit speaking task,feature extraction and speech recognition were also performed on the HD EMG signals for automatic speech recognition.In the vowel phonation task,the high-density sEMG energy map results showed that both facial and neck muscles are activated in the phonation task.The degree of activation and area of activated muscles both increased as the pitch gradually increased.At the same time,the pattern of the neck muscle activation areas was rather consistent with the air flow path during the phonation process.It could also be observed that the activity intensity of the neck muscles was significantly higher than that of the facial muscles,indicating that the neck muscles may play a more important role in the process of vocalization.The left-right symmetrical characteristics of the sEMG energy map indicated that the muscle activities had almost equal intensity on the right and left sides of the facial and neck muscles.In the task of speech recognition,the energy map calculated by the high-density sEMG signal showed that the muscle activity when speaking different digits showed significantly different patterns,and the EMG signal collected on the neck muscles could achieve better speech recognition accuracy.It indicated that the neck muscles contributed more than facial muscles in sEMG-based speech recognition.On this basis,this study further utilized the SFS algorithm to automatically select the best channel with the highest classification accuracy.The most important finding was that the best electrode selected by the SFS algorithm even showed great performance with a relatively small number of electrodes.For example,only 10 optimal electrodes were needed to achieve a classification accuracy close to 90%,which was significantly better than using all the 40 facial electrodes.The results of this section show that the selection and optimization of electrodes is very important for sEMG-based speech recognition.If there is no electrode optimization,for example,all electrodes are placed in less important locations(such as the face),the classification accuracy still stay as low as 73.6% even if the number of electrodes is as high as 40.The results of this study showed that the high-density sEMG technology can be used to realize the dynamic visualization of the swallowing and phonation process,to reveal the cooperative working mechanism of different muscles,and to achieve the objective evaluation of the swallowing or phonation functions.The findings are also useful for early diagnosis and targeted intervention of patients with swallowing or phonation disorders.