Merge pull request #3 from pablodz/dev

resampler: linear and decimation

examples/linear_resampler/main.go

@@ -0,0 +1,75 @@
+package main
+
+import (
+	"os"
+
+	"github.com/pablodz/sopro/pkg/audioconfig"
+	"github.com/pablodz/sopro/pkg/cpuarch"
+	"github.com/pablodz/sopro/pkg/encoding"
+	"github.com/pablodz/sopro/pkg/fileformat"
+	"github.com/pablodz/sopro/pkg/resampler"
+	"github.com/pablodz/sopro/pkg/transcoder"
+)
+
+func main() {
+
+	// Open the input file
+	in, err := os.Open("./internal/samples/v1_16b_16000.wav")
+	if err != nil {
+		panic(err)
+	}
+	defer in.Close()
+
+	// Create the output file
+	out, err := os.Create("./internal/samples/v1_16b_8000.wav")
+	if err != nil {
+		panic(err)
+	}
+	defer out.Close()
+
+	// create a transcoder
+	t := &transcoder.Transcoder{
+		MethodR: resampler.LINEAR_INTERPOLATION,
+		SourceConfigs: transcoder.TranscoderAudioConfig{
+			Endianness: cpuarch.LITTLE_ENDIAN,
+		},
+		TargetConfigs: transcoder.TranscoderAudioConfig{
+			Endianness: cpuarch.LITTLE_ENDIAN,
+		},
+		SizeBuffer: 1024,
+		Verbose:    true,
+	}
+
+	// Transcode the file
+	err = t.Wav2Wav(
+		&transcoder.AudioFileIn{
+			Data: in,
+			AudioFileGeneral: transcoder.AudioFileGeneral{
+				Format: fileformat.AUDIO_WAV,
+				Config: audioconfig.WavConfig{
+					BitDepth:   16,
+					Channels:   1,
+					Encoding:   encoding.SPACE_LINEAR, // ulaw is logarithmic
+					SampleRate: 16000,
+				},
+			},
+		},
+		&transcoder.AudioFileOut{
+			Data: out,
+			AudioFileGeneral: transcoder.AudioFileGeneral{
+				Format: fileformat.AUDIO_WAV,
+				Config: audioconfig.WavConfig{
+					BitDepth:   16,
+					Channels:   1,
+					Encoding:   encoding.SPACE_LINEAR,
+					SampleRate: 8000,
+				},
+			},
+		},
+	)
+
+	if err != nil {
+		panic(err)
+	}
+
+}

examples/ulaw2wav_logpcm/main.go

@@ -29,7 +29,7 @@ func main() {
 
 	// create a transcoder
 	t := &transcoder.Transcoder{
-		Method: method.BIT_TABLE,
+		MethodT: method.BIT_LOOKUP_TABLE,
 		SourceConfigs: transcoder.TranscoderAudioConfig{
 			Endianness: cpuarch.LITTLE_ENDIAN,
 		},

examples/ulaw2wav_lpcm/main.go

@@ -29,7 +29,7 @@ func main() {
 
 	// create a transcoder
 	t := &transcoder.Transcoder{
-		Method: method.BIT_TABLE,
+		MethodT: method.BIT_LOOKUP_TABLE,
 		SourceConfigs: transcoder.TranscoderAudioConfig{
 			Endianness: cpuarch.LITTLE_ENDIAN,
 		},

pkg/method/method.go

@@ -3,6 +3,13 @@ package method
 const (
 	NOT_FILLED            = (iota - 1) // Not filled
 	BIT_SHIFT                          // Bit shift
-	BIT_TABLE                          // Bit table, use a slice to store the values
+	BIT_LOOKUP_TABLE                   // Bit table, use a slice to store the values
 	BIT_ADVANCED_FUNCTION              // Advanced function, use a function to calculate and return the values
 )
+
+var METHODS = map[int]string{
+	NOT_FILLED:            "NOT_FILLED",
+	BIT_SHIFT:             "BIT_SHIFT",
+	BIT_LOOKUP_TABLE:      "BIT_LOOKUP_TABLE",
+	BIT_ADVANCED_FUNCTION: "BIT_ADVANCED_FUNCTION",
+}

pkg/downsampler/downsampler.go → pkg/resampler/downsampler.go

@@ -1,4 +1,4 @@
-package downsampler
+package resampler
 
 import "errors"
 

pkg/downsampler/downsampler_test.go → pkg/resampler/downsampler_test.go

@@ -1,4 +1,4 @@
-package downsampler
+package resampler
 
 import (
 	"reflect"

pkg/resampler/linear.go

@@ -0,0 +1,25 @@
+package resampler
+
+func LinearInterpolation[T int16 | int32 | int64 | int | byte](data []T, ratio float64) ([]T, error) {
+
+	// Calculate the length of the resampled data slice.
+	resampledLength := int(float64(len(data)) / ratio)
+
+	// Preallocate the resampled data slice with the correct size.
+	resampledData := make([]T, resampledLength)
+
+	// Iterate over the original data, interpolating new samples as necessary to
+	// resample the data at the target sample rate.
+	for i := 0; i < len(data)-1; i++ {
+		// Calculate the interpolated value between the current and next samples.
+		interpolatedValue := T((float64(data[i]) + float64(data[i+1])) / 2)
+		resampledData[int(float64(i)/ratio)] = interpolatedValue
+
+		// Skip the next sample if necessary.
+		if ratio > 1.0 {
+			i += int(ratio) - 1
+		}
+	}
+
+	return resampledData, nil
+}

pkg/resampler/linear_test.go

@@ -0,0 +1,48 @@
+package resampler
+
+import "testing"
+
+func TestLinearInterpolation(t *testing.T) {
+	data := []int16{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+	expected := []int16{1, 3, 5, 7, 9}
+	ratio := float64(16000) / float64(8000)
+	resampledData, err := LinearInterpolation(data, ratio)
+	if err != nil {
+		t.Errorf("Expected no error, got %v", err)
+	}
+	if !testEq(resampledData, expected) {
+		t.Errorf("Expected %v, got %v", expected, resampledData)
+	}
+}
+
+func TestLinearInterpolation2(t *testing.T) {
+	data := []int16{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
+	expected := []int16{1, 0, 2, 0, 3, 0, 4, 0, 5, 0, 6, 0, 7, 0, 8, 0, 9, 0, 0, 0}
+	ratio := float64(8000) / float64(16000)
+	resampledData, err := LinearInterpolation(data, ratio)
+	if err != nil {
+		t.Errorf("Expected no error, got %v", err)
+	}
+	if !testEq(resampledData, expected) {
+		t.Errorf("Expected %v, got %v", expected, resampledData)
+	}
+}
+
+// testEq is a helper function to compare two slices of int16 values.
+func testEq[T int16 | int32 | int64 | int](a, b []T) bool {
+	if a == nil && b == nil {
+		return true
+	}
+	if a == nil || b == nil {
+		return false
+	}
+	if len(a) != len(b) {
+		return false
+	}
+	for i := range a {
+		if a[i] != b[i] {
+			return false
+		}
+	}
+	return true
+}

pkg/upsampler/neuronal_networks.go → pkg/resampler/neuronal_networks.go

@@ -1,3 +1,3 @@
-package upsampler
+package resampler
 
 // TODO: Add externals connection to infer from an endpoint

pkg/resampler/resample.go

@@ -0,0 +1,13 @@
+package resampler
+
+const (
+	NOT_FILLED               = (iota - 1) // Not filled
+	LINEAR_INTERPOLATION                  // Linear interpolation
+	POLYNOMIAL_INTERPOLATION              // Polynomial interpolation
+)
+
+var RESAMPLER_METHODS = map[int]string{
+	NOT_FILLED:               "NOT_FILLED",
+	LINEAR_INTERPOLATION:     "LINEAR_INTERPOLATION",
+	POLYNOMIAL_INTERPOLATION: "POLYNOMIAL_INTERPOLATION",
+}

pkg/upsampler/upsampler.go → pkg/resampler/upsampler.go

@@ -1,3 +1,3 @@
-package upsampler
+package resampler
 
 // TODO: Add different upsampling methods

pkg/transcoder/models_transcoder.go

@@ -1,7 +1,8 @@
 package transcoder
 
 type Transcoder struct {
-	Method                int                   // the method of transcoding (e.g. 1, 2, 3, etc.)
+	MethodT               int                   // the method of transcoding (e.g. 1, 2, 3, etc.)
+	MethodR               int                   // the method of resampling (e.g. 1, 2, 3, etc.)
 	MethodAdvancedConfigs interface{}           // the specific configuration options for the transcoding method
 	SizeBuffer            int                   // the size of the buffer to read from the input file. Default is 1024
 	SourceConfigs         TranscoderAudioConfig // the source configuration

pkg/transcoder/mulaw2wav.go

@@ -15,9 +15,6 @@ import (
 	"golang.org/x/term"
 )
 
-var WIDTH_TERMINAL = 80
-var HEIGHT_TERMINAL = 30
-
 func init() {
 
 	err := error(nil)
@@ -27,7 +24,6 @@ func init() {
 	}
 }
 
-// TODO: split functions for different sizes of files
 // Transcode an ulaw file to a wav file (large files supported)
 // https://raw.githubusercontent.com/corkami/pics/master/binary/WAV.png
 // http://www.topherlee.com/software/pcm-tut-wavformat.html

pkg/transcoder/mulaw2wavlogpcm.go

@@ -22,7 +22,6 @@ func init() {
 	}
 }
 
-// TODO: split functions for different sizes of files
 // Transcode an ulaw file to a wav file (large files supported)
 // https://raw.githubusercontent.com/corkami/pics/master/binary/WAV.png
 // http://www.topherlee.com/software/pcm-tut-wavformat.html

pkg/transcoder/resampling_general.go

@@ -0,0 +1,70 @@
+package transcoder
+
+import (
+	"fmt"
+	"io"
+	"sync"
+
+	"github.com/pablodz/sopro/pkg/audioconfig"
+	"github.com/pablodz/sopro/pkg/resampler"
+)
+
+var doOnceResampling sync.Once
+
+func ResampleBytes(in *AudioFileIn, out *AudioFileOut, transcoder *Transcoder) error {
+
+	bitsProcessed, err := differentSampleRate(in, out, transcoder)
+	if err != nil {
+		return err
+	}
+	transcoder.Println("Transcoding done:", bitsProcessed, "bits processed")
+
+	return nil
+}
+
+func differentSampleRate(in *AudioFileIn, out *AudioFileOut, transcoder *Transcoder) (int, error) {
+	sizeBuff := 1024 // max size, more than that would be too much
+	if transcoder.SizeBuffer > 0 {
+		sizeBuff = transcoder.SizeBuffer
+	}
+	nTotal := 0
+	sampleRateIn := in.Config.(audioconfig.WavConfig).SampleRate
+	sampleRateOut := out.Config.(audioconfig.WavConfig).SampleRate
+	ratio := float64(sampleRateIn) / float64(sampleRateOut)
+
+	bufIn := make([]byte, sizeBuff)             // input buffer
+	bufOut := make([]byte, sizeBuff*int(ratio)) // output buffer
+	transcoder.Println("ratio", ratio, "ratioInt", int(ratio))
+	for {
+		n, err := in.Reader.Read(bufIn)
+		if err != nil && err != io.EOF {
+			return nTotal, fmt.Errorf("error reading input file: %v", err)
+		}
+		if n == 0 {
+			break
+		}
+		bufIn = bufIn[:n]
+		// buf2 is different size than buf
+		bufOut, _ = resampler.LinearInterpolation(bufIn, ratio) // IMPORTANT:buf cut to n bytes
+		out.Length += len(bufOut)
+		if _, err = out.Writer.Write(bufOut); err != nil {
+			return nTotal, fmt.Errorf("error writing output file: %v", err)
+		}
+		nTotal += n
+
+		doOnceResampling.Do(func() {
+			transcoder.Println("[Transcoder] Transcoding data - sample of the first 4 bytes (hex)")
+			onlyNFirst := 8
+			transcoder.Println(
+				"[OLD]", fmt.Sprintf("% 2x", bufIn[:onlyNFirst]),
+				"\n[NEW]", fmt.Sprintf("% 2x", bufOut[:onlyNFirst/2]),
+			)
+			transcoder.Println("[Transcoder] Transcoding data - sample of the first 4 bytes (decimal)")
+			transcoder.Println(
+				"[OLD]", fmt.Sprintf("%3d", bufIn[:onlyNFirst]),
+				"\n[NEW]", fmt.Sprintf("%3d", bufOut[:onlyNFirst/2]),
+			)
+		})
+	}
+	return nTotal, nil
+}

pkg/transcoder/router.go

@@ -5,8 +5,13 @@ import (
 
 	"github.com/pablodz/sopro/pkg/audioconfig"
 	"github.com/pablodz/sopro/pkg/encoding"
+	"github.com/pablodz/sopro/pkg/method"
+	"github.com/pablodz/sopro/pkg/resampler"
 )
 
+var WIDTH_TERMINAL = 80
+var HEIGHT_TERMINAL = 30
+
 const ErrUnsupportedConversion = "unsupported conversion"
 
 func (t *Transcoder) Mulaw2Wav(in *AudioFileIn, out *AudioFileOut) error {
@@ -15,9 +20,13 @@ func (t *Transcoder) Mulaw2Wav(in *AudioFileIn, out *AudioFileOut) error {
 	outSpace := out.Config.(audioconfig.WavConfig).Encoding
 
 	switch {
-	case inSpace == encoding.SPACE_LOGARITHMIC && outSpace == encoding.SPACE_LINEAR:
+	case t.MethodT != method.BIT_LOOKUP_TABLE &&
+		inSpace == encoding.SPACE_LOGARITHMIC &&
+		outSpace == encoding.SPACE_LINEAR:
 		return mulaw2WavLpcm(in, out, t)
-	case inSpace == encoding.SPACE_LOGARITHMIC && outSpace == encoding.SPACE_LOGARITHMIC:
+	case t.MethodT != method.BIT_LOOKUP_TABLE &&
+		inSpace == encoding.SPACE_LOGARITHMIC &&
+		outSpace == encoding.SPACE_LOGARITHMIC:
 		return mulaw2WavLogpcm(in, out, t)
 	default:
 		return fmt.Errorf(
@@ -29,3 +38,28 @@ func (t *Transcoder) Mulaw2Wav(in *AudioFileIn, out *AudioFileOut) error {
 
 	}
 }
+
+func (t *Transcoder) Wav2Wav(in *AudioFileIn, out *AudioFileOut) error {
+
+	inSpace := in.Config.(audioconfig.WavConfig).Encoding
+	outSpace := out.Config.(audioconfig.WavConfig).Encoding
+
+	switch {
+	case t.MethodR == resampler.LINEAR_INTERPOLATION &&
+		inSpace == encoding.SPACE_LINEAR &&
+		outSpace == encoding.SPACE_LINEAR:
+		return wavLpcm2wavLpcm(in, out, t)
+	case t.MethodR == resampler.LINEAR_INTERPOLATION &&
+		inSpace == encoding.SPACE_LOGARITHMIC &&
+		outSpace == encoding.SPACE_LOGARITHMIC:
+		fallthrough
+	default:
+		return fmt.Errorf(
+			"%s: %s -> %s",
+			ErrUnsupportedConversion,
+			encoding.ENCODINGS[inSpace],
+			encoding.ENCODINGS[outSpace],
+		)
+
+	}
+}