Merge pull request #11 from arjo129/3.0-workingBranch

Release V2.1

README.md

@@ -1,6 +1,6 @@
 uSpeech library
 ==============
-
+This  is branch is the 3.0 release sandbox. It is currently vaporware unless you wish to contribute.
 The uSpeech library provides an interface for voice recognition using the Arduino. It currently produces phonemes, often the library will produce junk phonemes. Please bare with it for the time being. A noise removal function is underway.
 ##Minimum Requirements ##
 The library is quite intensive on the processor. Each sample collection takes about 3.2 milliseconds so pay close attention to the time. The library has been tested on the Arduino Uno (ATMega32). Each signal object uses up 160bytes. No real time scheduler should be used with this.

installation/examples/Debug_uSpeech/Debug_uSpeech.ino

@@ -0,0 +1,42 @@
+#include <uspeech.h>
+signal voice(A0);
+char option = 0;
+void setup(){
+  voice.calibrate();
+  Serial.begin(9600);
+}
+void loop(){
+  voice.sample();
+  char c=voice.getPhoneme();
+  if(option==0){
+    Serial.println("uSpeech debug tool--------");
+    Serial.println(" Press 'a' if you wish to calibrate/check the f algorithm");
+    Serial.println(" Press 'b' if you wish to calibrate/check the getPhoneme");
+    Serial.println(" Press 'c' if you wish to calibrate/check the volume of your microphone");
+    Serial.println(" Press 'd' if you wish to calibrate/check the coeff");
+    Serial.println(" Press 'e' if you wish to calibrate/check the vowel detector");   
+    option = 1;
+  }
+  if(option==1){
+   if (Serial.available() > 0) {
+                // read the incoming byte:
+                option = Serial.read();
+   }
+  }
+  if(option=='a'){
+    Serial.println(voice.micPower);
+  }
+  if(option=='b'){
+    Serial.println(c);
+  }
+  if(option=='c'){
+    Serial.println(voice.power());
+  }
+  if(option=='d'){
+  	Serial.println(voice.testCoeff);
+  }
+  if(option=='e'){
+
+    Serial.println(voice.vowelRatio);
+  }
+}

keywords.txt

@@ -14,6 +14,7 @@ getPhoneme	KEYWORD2
 snr		KEYWORD2
 sample		KEYWORD2
 calibrate	KEYWORD2
+power	       KEYWORD2
 #######################################
 # Instances (KEYWORD2)
 #######################################

phoneme.cpp

@@ -4,9 +4,12 @@
 */
 char signal::getPhoneme(){
 	sample();
-	if(power()>SILENCE){
+    unsigned int pp =power();
+	if(pp>SILENCE){
+
 		//Low pass filter for noise removal
-		int k = complexity(power()); 
+		int k = complexity(pp);
+
 		overview[6] = overview[5];
 		overview[5] = overview[4];
 		overview[4] = overview[3];
@@ -21,17 +24,19 @@ char signal::getPhoneme(){
 			f++;
 		}
 		coeff /= 7;
+        testCoeff = coeff;
 		//Serial.println(coeff); //Use this for debugging
 #if F_DETECTION > 0
         micPower = 0.05 * maxPower() + (1 - 0.05) * micPower;
         //Serial.println(micPower)//If you are having trouble with fs
-
-        if (micPower > 37/*Replace this value (37) with your own*/) {
+        if (micPower > F_CONSTANT/*Use the header file to change this*/) {
             return 'f';
         }
 #endif
-	//Twiddle with the numbers here if your getting false triggers
+        zeroCrossingSearch();
+    //Twiddle with the numbers here if your getting false triggers
 	//This is the main recognizer part
+	//Todo: use move values to header file
 		if(coeff<30 && coeff>20){
 			return 'u';
 		}
@@ -65,18 +70,8 @@ char signal::getPhoneme(){
 		}
 	}
 	else{
+        micPower = 0;
 		return ' ';
 	}
 }
-void signal::formantAnal(){
-	int i = 0;
-	int k = 0;
-	while(i<18){
-		if((long)(filters[i]-filters[i-1]) > 0 & (long)(filters[i]-filters[i+1]) < 0){
-			if(k < 3){
-				formants[k] = i;
-			}
-		}
-		i++;
-	}
-}
+

signal.cpp

@@ -75,9 +75,12 @@ unsigned int signal::maxPower(){
     while (i<32){
         if(max<abs(arr[i])){
             max = abs(arr[i]);
+            maxPos = i;
         }
         i++;
+        avgPower+=arr[i];
     }
+    avgPower /= 32;
     return max;
 }
 int signal::snr(int power){
@@ -89,3 +92,27 @@ int signal::snr(int power){
 	}
 	return sqrt(j/mean)/power;
 }
+void signal::zeroCrossingSearch(){
+	int i=maxPos;
+	int prev = arr[i];
+	int upper = 0;
+	int lower = 0;
+	while (i<32){
+		prev = arr[i]-avgPower;
+		if(prev<0){
+			upper = i;
+			i = 33; //Break out of loop
+		}
+		i++;
+	}
+	i=maxPos;
+	while (i>0){
+		prev = arr[i]-avgPower;
+		if(prev<0){
+			lower = i;
+			i = 0; //Break out of loop
+		}
+		i--;
+	}
+	vowelRatio = (upper-i)*100/lower-i;
+}

uspeech.h

@@ -1,5 +1,5 @@
 /*
- uspeech v.1.9.3 (2.0betaRC2)
+ uspeech v.3.0Alpha
  2012 Arjo Chakravarty
  
  uspeech is a library that allows sounds to be classified into certain phonemes
@@ -11,12 +11,14 @@
 
 #include "Arduino.h"
 #include <math.h>
-#define SILENCE 92
+#define SILENCE 1500
 #define F_DETECTION 3
+#define F_CONSTANT 380
 class signal{
 public:
-	int arr[32];
+	int arr[32],avgPower;
 	int calib;
+    int testCoeff;
 	signal(int port);
     int micPower;
 	void sample();
@@ -26,19 +28,18 @@ class signal{
 	unsigned long fpowerex(int sum, int xtra); //Todo: Remove
 	int snr(int power);
 	void calibrate();
-	unsigned long filters[18], formants[3], intoned; //Todo: Remove
 	unsigned int overview[7];
-	char getPhoneme(); 
-	void debugPrintFilter(); //Todo: Remove
-	void voiceFormants(); //Todo: Remove
-	//void lowPass(int freq); Todo: Implement moving average low pass filter.
+	char getPhoneme(); //void lowPass(int freq); Todo: Implement moving average low pass filter.
     int goertzel(int freq);
-
+    int vowelRatio;
+    void zeroCrossingSearch();
 private:
 	int pin;
 	int mil;
+    int maxPos;
 	bool silence;
 	void formantAnal(); //Todo: Remove
+
 };