Engine sound generating system based on string sound source, and engine sound generating method
An engine sound generating system, positioned in a moving object, includes a repeating section selection unit for recognizing at least one vowel section from an input sound source, and selecting a repeating section to be repeated based on the vowel section; a repetitive waveform generating unit for generating a repetitive waveform based on the selected repeating section; and an engine sound output unit for generating and outputting an engine sound by interpolating the generated repetitive waveform.
1 . An engine sound generating system based on a string sound source, positioned in a moving object, the system comprising:
a repeating section selection unit configured to recognize a vowel section from an input sound source and to select a repeating section to be repeated based on the vowel section;
a repetitive waveform generating unit configured to generate a repetitive waveform based on the selected repeating section; and
an engine sound output unit configured to generate and output an engine sound by interpolating the repetitive waveform.
2 . The system of claim 1 , wherein:
the repeating section selection unit includes a string recognition unit configured to recognize a string from the input sound source, and
the string recognition unit is configured to calculate a probability that a waveform of the input sound source at a predetermined time point is to be classified into a predetermined letter.
3 . The system of claim 2 , wherein the string recognition unit is configured to extract a feature from the input sound source, to encode the extracted feature, and to classify the encoded feature into the predetermined letter and a word boundary.
4 . The system of claim 3 , wherein the string recognition unit is configured to:
match the predetermined letter for each extracted feature, and recognize the string by merging identical consecutive letters with each other.
5 . The system of claim 2 , wherein the repeating section selection unit is configured to output a boundary of each letter in the waveform based on the string.
6 . The system of claim 5 , wherein the repeating section selection unit is configured to:
recognize the vowel section by calculating a path having a maximum value among probabilities that the waveform at a specific time point maintains a current character or advances to a next character, and
match the string and the waveform with each other.
7 . The system of claim 1 , wherein the repeating section selection unit is configured to:
classify the vowel section for each period to calculate average signal power and a frequency response, and select a section, in which changes in the average signal power and the frequency response are minimized, as the selected repeating section.
8 . The system of claim 1 , wherein the repetitive waveform generating unit is configured to generate the repetitive waveform by interpolating a sample waveform having a period in units of real numbers.
9 . The system of claim 1 , wherein the engine sound output unit is configured to:
generate the engine sound when at least one of (i) an output of an accelerator position sensor (APS) positioned in the moving object and (ii) revolutions per minute (RPM) of the moving object, is rapidly increased, and generate the engine sound by interpolating the repetitive waveform while increasing a pitch scale factor of the repetitive waveform as the at least one of the output of the APS and the RPM of the moving object is rapidly increased.
10 . The system of claim 9 , wherein the engine sound output unit is configured to modulate a last part of a predetermined length of the input sound source based on a pitch of the engine sound, and then to output the engine sound when it is determined that an acceleration is saturated based on either the output of the APS or RPM of the moving object.
11 . An engine sound generating method based on a string sound source of an engine sound generating system which is positioned in a moving object, the engine sound generating unit including a repeating section selection unit, a repetitive waveform generating unit, and an engine sound output unit, the method comprising:
(a) recognizing, by the repeating section selection unit, a vowel section from an input sound source, and selecting a repeating section to be repeated based on the recognized vowel section;
(b) generating, by the repetitive waveform generating unit, a repetitive waveform based on the selected repeating section; and
(c) generating and outputting, by the engine sound output unit, an engine sound by interpolating the repetitive waveform.
12 . The method of claim 11 , wherein recognizing the vowel section further includes:
(a-1) recognizing a string from the input sound source;
(a-2) outputting a boundary of each letter in the waveform based on the recognized string, and
(a-3) using the recognized string from (a-1) for determining probability that a waveform of the input sound source at a predetermined time point is to be classified into a predetermined letter.
13 . The method of claim 12 , wherein, from (a-1), a feature is extracted from the input sound source, the extracted feature is encoded, and the encoded feature is classified into the predetermined letter and a word boundary.
14 . The method of claim 13 , wherein the predetermined letter is matched for each extracted feature, and the string is recognized by merging identical consecutive letters with each other.
15 . The method of claim 12 , wherein, from (a-2), the vowel section is recognized by calculating a path having a maximum value among probabilities that the waveform at a specific time point maintains a current character or advances to a next character, and matching the string and the waveform with each other.
16 . The method of claim 15 , wherein recognizing the vowel section (a-2) further includes:
classifying the recognized vowel section for each period to calculate average signal power and a frequency response, and selecting a section in which changes in the average signal power and the frequency response are minimized as the repeating section.
17 . The method of claim 11 , wherein the repetitive waveform is generated by interpolating a sample waveform having a period in units of real numbers.
18 . The method of claim 11 , wherein generating and outputting the engine sound further includes:
generating the engine sound when at least one of output of an accelerator position sensor (APS) positioned in the moving object or revolutions per minute (RPM) of the moving object is rapidly increased, and by interpolating the repetitive waveform while increasing a pitch scale factor of the repetitive waveform as the at least one of the output of the APS and the RPM of the moving object is rapidly increased.
19 . The method of claim 18 , wherein a last part of a predetermined length of the input sound source is modulated based on an increased pitch of the engine sound, and the engine sound is then output when it is determined that an acceleration is saturated based on either the output of the APS or the RPM of the moving object.