pcm: Added test and command for mono conversion. Also made resampleing use alsa.Buffer

audio/pcm/pcm.go

@@ -7,35 +7,33 @@ import (
 	"github.com/yobert/alsa"
 )
 
-// Resample resamples pcm data (inPcm) from 'fromRate' Hz to 'toRate' Hz and returns the resulting pcm.
+// Resample resamples pcm data from fromBuf to 'toRate' Hz and returns the resulting pcm.
-// If an error occurs, an error will be returned along with the original audio data
+// If an error occurs, an error will be returned along with the original fromBuf's data
-// 	- channels: number of channels
-// 	- bitDepth: number of bits in single sample
 // Notes:
-// 	- Input and output is assumed to be Little Endian.
+// 	- Currently only downsampling is implemented and fromBuf's rate must be divisible by toRate or an error will occur.
-// 	- Currently only downsampling is possible and fromRate must be divisible by toRate or an error will occur.
+// 	- If the number of bytes in fromBuf.Data is not divisible by the decimation factor (ratioFrom), the remaining bytes will
-// 	- If the number of bytes in 'inPcm' is not divisible by the decimation factor (ratioFrom), the remaining bytes will
 // 	  not be included in the result. Eg. input of length 480002 downsampling 6:1 will result in output length 80000.
-func Resample(inPcm []byte, fromRate, toRate, channels, bitDepth int) ([]byte, error) {
+func Resample(fromBuf alsa.Buffer, toRate int) ([]byte, error) {
+	fromRate := fromBuf.Format.Rate
 	if fromRate == toRate {
-		return inPcm, nil
+		return fromBuf.Data, nil
 	} else if fromRate < 0 {
-		return inPcm, fmt.Errorf("Unable to convert from: %v Hz", fromRate)
+		return fromBuf.Data, fmt.Errorf("Unable to convert from: %v Hz", fromRate)
 	} else if toRate < 0 {
-		return inPcm, fmt.Errorf("Unable to convert to: %v Hz", toRate)
+		return fromBuf.Data, fmt.Errorf("Unable to convert to: %v Hz", toRate)
 	}
 
 	// The number of bytes in a sample.
 	var sampleLen int
-	switch bitDepth {
+	switch fromBuf.Format.SampleFormat {
-	case 32:
+	case alsa.S32_LE:
-		sampleLen = 4 * channels
+		sampleLen = 4 * fromBuf.Format.Channels
-	case 16:
+	case alsa.S16_LE:
-		sampleLen = 2 * channels
+		sampleLen = 2 * fromBuf.Format.Channels
 	default:
-		return inPcm, fmt.Errorf("Unhandled bitDepth: %v, must be 16 or 32", bitDepth)
+		return fromBuf.Data, fmt.Errorf("Unhandled ALSA format: %v", fromBuf.Format.SampleFormat)
 	}
-	inPcmLen := len(inPcm)
+	inPcmLen := len(fromBuf.Data)
 
 	// Calculate sample rate ratio ratioFrom:ratioTo.
 	rateGcd := gcd(toRate, fromRate)
@@ -44,32 +42,30 @@ func Resample(inPcm []byte, fromRate, toRate, channels, bitDepth int) ([]byte, e
 
 	// ratioTo = 1 is the only number that will result in an even sampling.
 	if ratioTo != 1 {
-		return inPcm, fmt.Errorf("%v:%v is an unhandled from:to rate ratio. must be n:1 for some rate n", ratioFrom, ratioTo)
+		return fromBuf.Data, fmt.Errorf("%v:%v is an unhandled from:to rate ratio. must be n:1 for some rate n", ratioFrom, ratioTo)
 	}
 
 	newLen := inPcmLen / ratioFrom
 	result := make([]byte, 0, newLen)
 
-	// For each new sample to be generated, loop through the respective 'ratioFrom' samples in 'inPcm' to add them
+	// For each new sample to be generated, loop through the respective 'ratioFrom' samples in 'fromBuf.Data' to add them
 	// up and average them. The result is the new sample.
 	for i := 0; i < newLen/sampleLen; i++ {
 		var sum int
 		for j := 0; j < ratioFrom; j++ {
-			switch bitDepth {
+			switch fromBuf.Format.SampleFormat {
-			case 32:
+			case alsa.S32_LE:
-				sum += int(int32(binary.LittleEndian.Uint32(inPcm[(i*ratioFrom*sampleLen)+(j*sampleLen) : (i*ratioFrom*sampleLen)+((j+1)*sampleLen)])))
+				sum += int(int32(binary.LittleEndian.Uint32(fromBuf.Data[(i*ratioFrom*sampleLen)+(j*sampleLen) : (i*ratioFrom*sampleLen)+((j+1)*sampleLen)])))
-			case 16:
+			case alsa.S16_LE:
-				sum += int(int16(binary.LittleEndian.Uint16(inPcm[(i*ratioFrom*sampleLen)+(j*sampleLen) : (i*ratioFrom*sampleLen)+((j+1)*sampleLen)])))
+				sum += int(int16(binary.LittleEndian.Uint16(fromBuf.Data[(i*ratioFrom*sampleLen)+(j*sampleLen) : (i*ratioFrom*sampleLen)+((j+1)*sampleLen)])))
-			default:
-				return inPcm, fmt.Errorf("Unhandled bitDepth: %v, must be 16 or 32", bitDepth)
 			}
 		}
 		avg := sum / ratioFrom
 		bAvg := make([]byte, sampleLen)
-		switch bitDepth {
+		switch fromBuf.Format.SampleFormat {
-		case 32:
+		case alsa.S32_LE:
 			binary.LittleEndian.PutUint32(bAvg, uint32(avg))
-		case 16:
+		case alsa.S16_LE:
 			binary.LittleEndian.PutUint16(bAvg, uint16(avg))
 		}
 		result = append(result, bAvg...)
@@ -81,11 +77,10 @@ func Resample(inPcm []byte, fromRate, toRate, channels, bitDepth int) ([]byte, e
 // the given stereo recording (ALSA buffer)
 // if an error occurs, an error will be returned along with the original stereo data.
 func StereoToMono(stereoBuf alsa.Buffer) ([]byte, error) {
-	bufChannels := stereoBuf.Format.Channels
+	if stereoBuf.Format.Channels == 1 {
-	if bufChannels == 1 {
 		return stereoBuf.Data, nil
-	} else if bufChannels != 2 {
+	} else if stereoBuf.Format.Channels != 2 {
-		return stereoBuf.Data, fmt.Errorf("Audio is not stereo or mono, it has %v channels", bufChannels)
+		return stereoBuf.Data, fmt.Errorf("Audio is not stereo or mono, it has %v channels", stereoBuf.Format.Channels)
 	}
 
 	var stereoSampleBytes int

audio/pcm/pcm_test.go

@@ -5,10 +5,12 @@ import (
 	"io/ioutil"
 	"log"
 	"testing"
+
+	"github.com/yobert/alsa"
 )
 
-// TestResample accepts an input pcm file (assumed to be mono and using 16-bit samples) and outputs a resampled pcm file.
+// TestResample tests the Resample function using a pcm file that contains audio of a freq. sweep.
-// Input and output file names can be specified as arguments.
+// The output of the Resample function is compared with a file containing the expected result.
 func TestResample(t *testing.T) {
 	inPath := "../../../test/test-data/av/input/sweep_400Hz_20000Hz_-3dBFS_5s_48khz.pcm"
 	expPath := "../../../test/test-data/av/output/sweep_400Hz_20000Hz_resampled_48to8kHz.pcm"
@@ -19,8 +21,19 @@ func TestResample(t *testing.T) {
 		log.Fatal(err)
 	}
 
+	format := alsa.BufferFormat{
+		Channels:     1,
+		Rate:         48000,
+		SampleFormat: alsa.S16_LE,
+	}
+
+	buf := alsa.Buffer{
+		Format: format,
+		Data:   inPcm,
+	}
+
 	// Resample pcm.
-	resampled, err := Resample(inPcm, 48000, 8000, 1, 16)
+	resampled, err := Resample(buf, 8000)
 	if err != nil {
 		log.Fatal(err)
 	}
@@ -36,3 +49,44 @@ func TestResample(t *testing.T) {
 		t.Error("Resampled data does not match expected result.")
 	}
 }
+
+// TestStereoToMono tests the StereoToMono function using a pcm file that contains stereo audio.
+// The output of the StereoToMono function is compared with a file containing the expected mono audio.
+func TestStereoToMono(t *testing.T) {
+	inPath := "../../../test/test-data/av/input/stereo_DTMF_tones.pcm"
+	expPath := "../../../test/test-data/av/output/mono_DTMF_tones.pcm"
+
+	// Read input pcm.
+	inPcm, err := ioutil.ReadFile(inPath)
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	format := alsa.BufferFormat{
+		Channels:     2,
+		Rate:         44100,
+		SampleFormat: alsa.S16_LE,
+	}
+
+	buf := alsa.Buffer{
+		Format: format,
+		Data:   inPcm,
+	}
+
+	// Convert audio.
+	mono, err := StereoToMono(buf)
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	// Read expected mono pcm.
+	exp, err := ioutil.ReadFile(expPath)
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	// Compare result with expected.
+	if !bytes.Equal(mono, exp) {
+		t.Error("Converted data does not match expected result.")
+	}
+}

exp/pcm/resample/resample.go

@@ -7,23 +7,24 @@ import (
 	"log"
 
 	"bitbucket.org/ausocean/av/audio/pcm"
+	"github.com/yobert/alsa"
 )
 
 // This program accepts an input pcm file and outputs a resampled pcm file.
-// Input and output file names, to and from sample rates, channels and bit-depth can be specified as arguments.
+// Input and output file names, to and from sample rates, channels and sample format can be specified as arguments.
 func main() {
 	var inPath string
 	var outPath string
 	var from int
 	var to int
 	var channels int
-	var bitDepth int
+	var sf string
 	flag.StringVar(&inPath, "in", "data.pcm", "file path of input data")
 	flag.StringVar(&outPath, "out", "resampled.pcm", "file path of output")
 	flag.IntVar(&from, "from", 48000, "sample rate of input file")
 	flag.IntVar(&to, "to", 8000, "sample rate of output file")
 	flag.IntVar(&channels, "ch", 1, "number of channels in input file")
-	flag.IntVar(&bitDepth, "bd", 16, "bit depth of input file")
+	flag.StringVar(&sf, "sf", "S16_LE", "sample format of input audio, eg. S16_LE")
 	flag.Parse()
 
 	// Read pcm.
@@ -33,8 +34,29 @@ func main() {
 	}
 	fmt.Println("Read", len(inPcm), "bytes from file", inPath)
 
-	// Resample pcm.
+	var sampleFormat alsa.FormatType
-	resampled, err := pcm.Resample(inPcm, from, to, channels, bitDepth)
+	switch sf {
+	case "S32_LE":
+		sampleFormat = alsa.S32_LE
+	case "S16_LE":
+		sampleFormat = alsa.S16_LE
+	default:
+		log.Fatalf("Unhandled ALSA format: %v", sf)
+	}
+
+	format := alsa.BufferFormat{
+		Channels:     channels,
+		Rate:         from,
+		SampleFormat: sampleFormat,
+	}
+
+	buf := alsa.Buffer{
+		Format: format,
+		Data:   inPcm,
+	}
+
+	// Resample audio.
+	resampled, err := pcm.Resample(buf, to)
 	if err != nil {
 		log.Fatal(err)
 	}

exp/pcm/stereo-to-mono/stereo-to-mono.go

@@ -0,0 +1,63 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"io/ioutil"
+	"log"
+
+	"bitbucket.org/ausocean/av/audio/pcm"
+	"github.com/yobert/alsa"
+)
+
+// This program accepts an input pcm file and outputs a resampled pcm file.
+// Input and output file names, to and from sample rates, channels and sample format can be specified as arguments.
+func main() {
+	var inPath string
+	var outPath string
+	var sf string
+	flag.StringVar(&inPath, "in", "data.pcm", "file path of input data")
+	flag.StringVar(&outPath, "out", "mono.pcm", "file path of output")
+	flag.StringVar(&sf, "sf", "S16_LE", "sample format of input audio, eg. S16_LE")
+	flag.Parse()
+
+	// Read pcm.
+	inPcm, err := ioutil.ReadFile(inPath)
+	if err != nil {
+		log.Fatal(err)
+	}
+	fmt.Println("Read", len(inPcm), "bytes from file", inPath)
+
+	var sampleFormat alsa.FormatType
+	switch sf {
+	case "S32_LE":
+		sampleFormat = alsa.S32_LE
+	case "S16_LE":
+		sampleFormat = alsa.S16_LE
+	default:
+		log.Fatalf("Unhandled ALSA format: %v", sf)
+	}
+
+	format := alsa.BufferFormat{
+		Channels:     2,
+		SampleFormat: sampleFormat,
+	}
+
+	buf := alsa.Buffer{
+		Format: format,
+		Data:   inPcm,
+	}
+
+	// Convert audio.
+	mono, err := pcm.StereoToMono(buf)
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	// Save mono to file.
+	err = ioutil.WriteFile(outPath, mono, 0644)
+	if err != nil {
+		log.Fatal(err)
+	}
+	fmt.Println("Encoded and wrote", len(mono), "bytes to file", outPath)
+}