+ * This component implements the speech processor interface using the Azure + * Speech Service for speech recognition. + *
+ * + *+ * When the speech context is triggered, the recognizer begins streaming + * buffered frames to the API for recognition. Once the speech context becomes + * inactive, the recognizer raises a RECOGNIZE event along with the audio + * transcript. Unfortunately, the Azure Speech SDK currently doesn't return + * confidence values alongside transcripts, so confidence is always set to 1.0. + *
+ * + *+ * Use of the Azure Speech Service implies acceptance of Microsoft's license + * terms, which can be found + * + * here. + *
+ * + *+ * This pipeline component requires the following configuration properties: + *
+ *- * This class uses the Microsoft speech recognition service for automatic - * speech recognition. It implements the client side of the websocket protocol - * defined here. Transcripts and errors - * are delivered to the client asynchronously via the {@link Listener} - * interface. The following sample demonstrates how to configure and use the - * client. - *
- * - *
- * {@code
- * BingSpeechClient client = new BingSpeechClient.Builder()
- * .setApiKey("")
- * .setLocale("en-US")
- * .setSampleRate(16000)
- * .setListener(this)
- * .build();
- *
- * client.connect();
- *
- * client.beginAudio();
- * for (ByteBuffer frame: frames)
- * client.sendAudio(frame);
- * client.endAudio();
- *
- * client.disconnect();
- * }
- *
- */
-public class BingSpeechClient implements AutoCloseable {
- private static final int MESSAGE_BUFFER_SIZE = 8192;
- private static final String TOKEN_URL =
- "https://api.cognitive.microsoft.com/sts/v1.0/issueToken";
- private static final String TOKEN_HEADER_APIKEY =
- "Ocp-Apim-Subscription-Key";
- private static final String SOCKET_URL =
- "wss://speech.platform.bing.com"
- + "/speech/recognition/conversation/cognitiveservices/v1?language=%s";
-
- private final int sampleRate;
- private final String locale;
- private final Listener listener;
- private final OkHttpClient client;
- private final String token;
- private final ByteBuffer buffer;
- private WebSocket socket;
- private String requestId;
-
- BingSpeechClient(Builder builder) throws Exception {
- this.sampleRate = builder.sampleRate;
- this.locale = builder.locale;
- this.listener = builder.listener;
- this.buffer = ByteBuffer.allocateDirect(MESSAGE_BUFFER_SIZE);
- this.client = builder.getHttpClient();
-
- Request tokenRequest = new Request.Builder()
- .url(TOKEN_URL)
- .addHeader(TOKEN_HEADER_APIKEY, builder.apiKey)
- .post(RequestBody.create(null, new byte[0]))
- .build();
- Response tokenResponse = this.client
- .newCall(tokenRequest)
- .execute();
- if (tokenResponse.code() != 200)
- throw new Exception("authentication failed");
- this.token = tokenResponse.body().string();
- }
-
- /**
- * releases the resources associated with the speech client.
- */
- public void close() {
- disconnect();
- }
-
- /**
- * @return true if the websocket is currently connect, false otherwise
- */
- public boolean isConnected() {
- return this.socket != null;
- }
-
- /**
- * @return the request identifier of the current turn
- */
- public String getRequestId() {
- return this.requestId;
- }
-
- /**
- * establishes a websocket connection for speech recognition.
- */
- public void connect() {
- if (this.socket != null)
- throw new IllegalStateException();
-
- String connectionId = UUID.randomUUID().toString().replace("-", "");
- Request request = new Request.Builder()
- .url(String.format(SOCKET_URL, this.locale))
- .addHeader("Authorization", String.format("Bearer %s", this.token))
- .addHeader("X-ConnectionId", connectionId)
- .build();
- this.socket = this.client.newWebSocket(request, new SocketListener());
- }
-
- /**
- * disconnects the websocket.
- */
- public void disconnect() {
- if (this.socket != null) {
- try {
- this.socket.close(1000, "goodbye");
- } finally {
- this.socket = null;
- this.requestId = null;
- this.buffer.clear();
- }
- }
- }
-
- /**
- * begins audio transmission for a new conversation turn.
- * @throws Exception on error
- */
- public void beginAudio() throws Exception {
- if (this.socket == null)
- throw new IllegalStateException();
-
- this.requestId = UUID.randomUUID().toString().replace("-", "");
- sendConfig();
- sendHeader();
- }
-
- private void sendConfig() throws Exception {
- String header =
- "Path:speech.config\r\n"
- + String.format("X-RequestId:%s\r\n", this.requestId)
- + String.format("X-Timestamp:%s\r\n", now())
- + "Content-Type:application/json; charset=utf-8\r\n";
- String body = "{}";
- String message = String.format("%s\r\n%s", header, body);
- this.socket.send(message);
- }
-
- private void sendHeader() throws Exception {
- ByteBuffer message = ByteBuffer.allocateDirect(44)
- .order(ByteOrder.LITTLE_ENDIAN) // wave endian
- .put("RIFF".getBytes("US-ASCII")) // riff chunk id
- .putInt(0) // file size
- .put("WAVE".getBytes("US-ASCII")) // riff format
- .put("fmt ".getBytes("US-ASCII")) // begin format chunk
- .putInt(16) // format chunk size
- .putShort((short) 1) // format (1=PCM)
- .putShort((short) 1) // channel count
- .putInt(this.sampleRate) // sample rate
- .putInt(this.sampleRate * 2) // byte rate
- .putShort((short) 2) // block alignment
- .putShort((short) 16) // bits per sample
- .put("data".getBytes("US-ASCII")) // begin data chunk
- .putInt(0); // data chunk size
- sendAudio(message);
- flush();
- }
-
- /**
- * transmits an audio frame over the websocket.
- * @param frame the audio frame buffer to send
- * @throws Exception on error
- */
- public void sendAudio(ByteBuffer frame) throws Exception {
- if (this.socket == null)
- throw new IllegalStateException();
-
- if (frame == null || this.buffer.remaining() < frame.capacity())
- flush();
-
- if (this.buffer.position() == 0) {
- byte[] header = (
- "Path:audio\r\n"
- + String.format("X-RequestId:%s\r\n", this.requestId)
- + String.format("X-Timestamp:%s\r\n", now())
- + "Content-Type:audio/x-wav\r\n"
- ).getBytes("US-ASCII");
-
- this.buffer
- .order(ByteOrder.BIG_ENDIAN) // protocol endian
- .putShort((short) header.length) // message header length
- .put(header); // message header
- }
-
- if (frame != null) {
- frame.rewind();
- this.buffer.put(frame); // message body
- }
- }
-
- /**
- * sends an empty audio frame indicating the end of the turn.
- * @throws Exception on error
- */
- public void endAudio() throws Exception {
- if (this.socket == null)
- throw new IllegalStateException();
-
- sendAudio(null);
- flush();
- }
-
- private void flush() {
- if (this.buffer.position() > 0) {
- this.buffer.flip();
- this.socket.send(ByteString.of(this.buffer));
- this.buffer.clear();
- }
- }
-
- private String now() {
- SimpleDateFormat formatter =
- new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss.SSS'000Z'");
- formatter.setTimeZone(TimeZone.getTimeZone("UTC"));
- return formatter.format(new Date());
- }
-
- /** okhttp socket callback. */
- private class SocketListener extends WebSocketListener {
- private final Gson gson = new Gson();
- private final Listener listener = BingSpeechClient.this.listener;
-
- @Override
- public void onMessage(WebSocket s, String message) {
- String requestHeader = String.format(
- "X-RequestId:%s",
- BingSpeechClient.this.requestId
- );
- if (message.contains(requestHeader))
- if (message.contains("Path:speech.phrase"))
- onSpeechPhrase(message);
- }
-
- @Override
- public void onClosed(WebSocket s, int code, String reason) {
- if (code != 1000)
- this.listener.onError(
- new Exception(String.format("%d: %s", code, reason))
- );
- }
-
- @Override
- public void onFailure(WebSocket s, Throwable e, Response r) {
- this.listener.onError(e);
- }
-
- private void onSpeechPhrase(String message) {
- String body = message.split("\r\n\r\n", 2)[1];
- Map result = gson.fromJson(body, Map.class);
- String status = (String) result.get("RecognitionStatus");
- if (status.equals("Success"))
- this.listener.onSpeech((String) result.get("DisplayText"));
- else if (!status.equals("EndOfDictation"))
- this.listener.onSpeech("");
- }
- }
-
- /**
- * bing speech client builder.
- */
- public static class Builder {
- private String apiKey;
- private String locale;
- private int sampleRate;
- private Listener listener;
-
- /**
- * sets the microsoft azure api key for authentication.
- * @param value api key to configure
- * @return this
- */
- public Builder setApiKey(String value) {
- this.apiKey = value;
- return this;
- }
-
- /**
- * sets the country/language code string.
- * @param value locale code to configure
- * @return this
- */
- public Builder setLocale(String value) {
- this.locale = value;
- return this;
- }
-
- /**
- * sets the audio sample rate.
- * @param value sample rate to configure, in Hz
- * @return this
- */
- public Builder setSampleRate(int value) {
- this.sampleRate = value;
- return this;
- }
-
- /**
- * attaches the listener callback.
- * @param value event listener to configure
- * @return this
- */
- public Builder setListener(Listener value) {
- this.listener = value;
- return this;
- }
-
- /**
- * initializes a new speech client instance.
- * @return the constructed speech client
- * @throws Exception on error
- */
- public BingSpeechClient build() throws Exception {
- if (this.apiKey == null)
- throw new IllegalArgumentException("apiKey");
- if (this.locale == null)
- throw new IllegalArgumentException("locale");
- if (this.sampleRate == 0)
- throw new IllegalArgumentException("sampleRate");
- if (this.listener == null)
- throw new IllegalArgumentException("listener");
- return new BingSpeechClient(this);
- }
-
- OkHttpClient getHttpClient() {
- return new OkHttpClient.Builder().build();
- }
- }
-
- /**
- * speech client listener callback interface.
- */
- public interface Listener {
- /**
- * called when a speech transcription is received.
- * @param transcript the speech transcript, or "" if no speech
- * was detected
- */
- void onSpeech(String transcript);
-
- /**
- * called when a speech detection error occurred.
- * @param e the speech error
- */
- void onError(Throwable e);
- }
-}
diff --git a/src/main/java/io/spokestack/spokestack/microsoft/BingSpeechRecognizer.java b/src/main/java/io/spokestack/spokestack/microsoft/BingSpeechRecognizer.java
deleted file mode 100644
index 073823c..0000000
--- a/src/main/java/io/spokestack/spokestack/microsoft/BingSpeechRecognizer.java
+++ /dev/null
@@ -1,156 +0,0 @@
-package io.spokestack.spokestack.microsoft;
-
-import java.nio.ByteBuffer;
-
-import io.spokestack.spokestack.SpeechConfig;
-import io.spokestack.spokestack.SpeechProcessor;
-import io.spokestack.spokestack.SpeechContext;
-
-/**
- * microsoft bing speech recognizer
- *
- * - * This component implements the speech processor interface using the - * Bing Speech API for speech recognition. It uses the websocket protocol - * defined here to stream audio samples - * and receive recognition results asynchronously. - *
- * - *- * When the speech context is triggered, the recognizer begins streaming - * buffered frames to the API for recognition. Once the speech context - * becomes inactive, the recognizer raises a RECOGNIZE event along with the - * audio transcript. Unfortunately, the Bing Speech API currently doesn't - * return confidence values alongside transcripts, so confidence is always - * set to 1.0. - *
- * - *- * This pipeline component requires the following configuration properties: - *
- *+ * Azure Speech Service requires extra configuration, which must be added to the + * pipeline build process separately from this profile: + *
+ * + *+ * Wakeword detection requires configuration to locate the models used for + * classification; these properties must be set elsewhere: + *
+ * + *+ * Azure Speech Service requires the following properties: + *
+ * + *+ * Azure Speech Service requires extra configuration, which must be added to the + * pipeline build process separately from this profile: + *
+ * + *