A Machine Learning Assessment System for Spoken English Based on Linear Predictive Coding

Table 5

English speech recognition accuracy under different convolution kernel conditions.

Data sets	Maximum accuracy	Average accuracy rate	Standard deviation
Convolution kernel size
1	0.9491	0.9291	1.75e − 003
2	0.9663	0.9307	1.92e − 003
3	0.9318	0.9267	1.66e − 003
4	0.9491	0.9244	1.73e − 003
5	0.9265	0.9074	1.81e − 003
6	0.9214	0.9083	1.65e − 003
Convolution kernel size
1	0.9482	0.9286	1.77e − 003
2	0.9658	0.9301	1.96e − 003
3	0.9316	0.9259	1.73e − 003
4	0.9492	0.9238	1.72e − 003
5	0.9263	0.9066	1.82e − 003
6	0.9209	0.9078	1.66e − 003
Convolution kernel size
1	0.8471	0.8312	4.13e − 003
2	0.8642	0.8476	3.37e − 003
3	0.8225	0.8164	5.68e − 003
4	0.8366	0.8132	4.05e − 003
5	0.8389	0.8246	3.49e − 003
6	0.8217	0.8059	4.22e − 003
Convolution kernel size
1	0.7323	0.7191	8.24e − 003
2	0.6824	0.6162	7.93e − 003
3	0.6917	0.6694	7.27e − 003
4	0.7318	0.7161	7.13e − 003
5	0.6429	0.6225	6.83e − 003
6	0.6835	0.6634	6.65e − 003